id stringlengths 33 33 | doc_url stringlengths 33 33 | title stringlengths 13 211 | publication_date timestamp[s] | update_date timestamp[s] | authors stringlengths 6 16.6k | category_primary stringclasses 60
values | category_all stringlengths 5 132 | abstract stringlengths 123 1.99k | text stringlengths 190 2.08k | embeddings listlengths 768 768 |
|---|---|---|---|---|---|---|---|---|---|---|
http://arxiv.org/abs/2407.02333v1 | http://arxiv.org/pdf/2407.02333v1 | Why do LLaVA Vision-Language Models Reply to Images in English? | 2024-07-02T15:01:55 | 2024-07-02T15:01:55 | Musashi Hinck, Carolin Holtermann, Matthew Lyle Olson, Florian Schneider, Sungduk Yu, Anahita Bhiwandiwalla, Anne Lauscher, Shaoyen Tseng, Vasudev Lal | cs.CL | cs.CL, cs.CV | We uncover a surprising multilingual bias occurring in a popular class of
multimodal vision-language models (VLMs). Including an image in the query to a
LLaVA-style VLM significantly increases the likelihood of the model returning
an English response, regardless of the language of the query. This paper
investigates the... | Why do LLaVA Vision-Language Models Reply to Images in English?: We uncover a surprising multilingual bias occurring in a popular class of
multimodal vision-language models (VLMs). Including an image in the query to a
LLaVA-style VLM significantly increases the likelihood of the model returning
an English response, reg... | [
0.0238626525,
0.0259211697,
-0.0377753004,
0.0787944049,
-0.0077792443,
0.0071066665,
0.0316351466,
-0.0114007052,
0.0304484852,
-0.0600155629,
-0.0468650833,
-0.0692457482,
0.0268885754,
0.0197670907,
0.0701574832,
-0.0637963489,
0.0334584787,
-0.047486417,
-0.0435186699,
-0.0... |
http://arxiv.org/abs/2407.02328v1 | http://arxiv.org/pdf/2407.02328v1 | Efficient Sparse Attention needs Adaptive Token Release | 2024-07-02T14:58:44 | 2024-07-02T14:58:44 | Chaoran Zhang, Lixin Zou, Dan Luo, Min Tang, Xiangyang Luo, Zihao Li, Chenliang Li | cs.CL | cs.CL | In recent years, Large Language Models (LLMs) have demonstrated remarkable
capabilities across a wide array of text-centric tasks. However, their `large'
scale introduces significant computational and storage challenges, particularly
in managing the key-value states of the transformer, which limits their wider
applicab... | Efficient Sparse Attention needs Adaptive Token Release: In recent years, Large Language Models (LLMs) have demonstrated remarkable
capabilities across a wide array of text-centric tasks. However, their `large'
scale introduces significant computational and storage challenges, particularly
in managing the key-value sta... | [
0.0382155031,
0.0549579225,
0.0129420087,
0.0172246974,
-0.0147748711,
-0.0091385981,
-0.0237931088,
-0.0032814303,
-0.0131802158,
-0.0340889096,
-0.0329443142,
-0.055887714,
-0.0106045175,
0.0388200358,
0.0298196618,
-0.0180634744,
0.041260507,
-0.0429969691,
0.0055841724,
0.0... |
http://arxiv.org/abs/2407.02320v1 | http://arxiv.org/pdf/2407.02320v1 | Exploring the Role of Transliteration in In-Context Learning for Low-resource Languages Written in Non-Latin Scripts | 2024-07-02T14:51:20 | 2024-07-02T14:51:20 | Chunlan Ma, Yihong Liu, Haotian Ye, Hinrich Schütze | cs.CL | cs.CL, cs.AI | Decoder-only large language models (LLMs) excel in high-resource languages
across various tasks through few-shot or even zero-shot in-context learning
(ICL). However, their performance often does not transfer well to low-resource
languages, especially those written in non-Latin scripts. Inspired by recent
work that lev... | Exploring the Role of Transliteration in In-Context Learning for Low-resource Languages Written in Non-Latin Scripts: Decoder-only large language models (LLMs) excel in high-resource languages
across various tasks through few-shot or even zero-shot in-context learning
(ICL). However, their performance often does not t... | [
0.0051830183,
0.039340131,
0.0099420762,
0.0292994846,
-0.0171325821,
0.0629817098,
0.0616949499,
0.0274308827,
-0.03054264,
-0.0140524227,
-0.0517802909,
-0.0741262585,
0.046772778,
-0.0223302767,
0.0294397566,
-0.0345281437,
-0.0008922986,
-0.0672405884,
-0.0064162444,
-0.005... |
http://arxiv.org/abs/2407.02317v1 | http://arxiv.org/pdf/2407.02317v1 | Soft Language Prompts for Language Transfer | 2024-07-02T14:50:03 | 2024-07-02T14:50:03 | Ivan Vykopal, Simon Ostermann, Marián Šimko | cs.CL | cs.CL | Cross-lingual knowledge transfer, especially between high- and low-resource
languages, remains a challenge in natural language processing (NLP). This study
offers insights for improving cross-lingual NLP applications through the
combination of parameter-efficient fine-tuning methods. We systematically
explore strategie... | Soft Language Prompts for Language Transfer: Cross-lingual knowledge transfer, especially between high- and low-resource
languages, remains a challenge in natural language processing (NLP). This study
offers insights for improving cross-lingual NLP applications through the
combination of parameter-efficient fine-tuning... | [
0.028763121,
-0.0421449132,
-0.0397074446,
0.0406603776,
-0.0293930136,
0.00597568,
0.0157796796,
-0.0353974216,
0.0174681488,
-0.0017840648,
-0.055185318,
0.0198470652,
0.0208668448,
0.0023075608,
0.0129807945,
-0.0531151257,
0.0749453232,
-0.0604544953,
0.0154795265,
-0.01027... |
http://arxiv.org/abs/2407.02310v1 | http://arxiv.org/pdf/2407.02310v1 | Evaluating the Ability of LLMs to Solve Semantics-Aware Process Mining Tasks | 2024-07-02T14:44:49 | 2024-07-02T14:44:49 | Adrian Rebmann, Fabian David Schmidt, Goran Glavaš, Han van der Aa | cs.CL | cs.CL | The process mining community has recently recognized the potential of large
language models (LLMs) for tackling various process mining tasks. Initial
studies report the capability of LLMs to support process analysis and even, to
some extent, that they are able to reason about how processes work. This latter
property su... | Evaluating the Ability of LLMs to Solve Semantics-Aware Process Mining Tasks: The process mining community has recently recognized the potential of large
language models (LLMs) for tackling various process mining tasks. Initial
studies report the capability of LLMs to support process analysis and even, to
some extent,... | [
0.0094749173,
0.0293672346,
-0.0295822136,
0.0543893203,
-0.0315968506,
-0.0108434008,
-0.0230436549,
-0.0095320484,
-0.0374257714,
-0.021007454,
0.0054631671,
-0.0009461112,
-0.0104477201,
0.0514391735,
-0.0273789801,
0.0138340611,
0.044547718,
-0.0286970213,
0.048946958,
0.06... |
http://arxiv.org/abs/2407.02302v1 | http://arxiv.org/pdf/2407.02302v1 | Towards Human Understanding of Paraphrase Types in ChatGPT | 2024-07-02T14:35:10 | 2024-07-02T14:35:10 | Dominik Meier, Jan Philip Wahle, Terry Ruas, Bela Gipp | cs.CL | cs.CL, I.2.7 | Paraphrases represent a human's intuitive ability to understand expressions
presented in various different ways. Current paraphrase evaluations of language
models primarily use binary approaches, offering limited interpretability of
specific text changes. Atomic paraphrase types (APT) decompose paraphrases into
differe... | Towards Human Understanding of Paraphrase Types in ChatGPT: Paraphrases represent a human's intuitive ability to understand expressions
presented in various different ways. Current paraphrase evaluations of language
models primarily use binary approaches, offering limited interpretability of
specific text changes. Atom... | [
0.1040409133,
0.0340980291,
-0.0295291934,
0.0299925413,
-0.0681817308,
0.0047472692,
-0.0275105499,
0.0231127758,
-0.0077399733,
-0.0693191662,
-0.0151125481,
0.0280231442,
0.0085145785,
-0.0266337264,
0.0197840296,
-0.0954611301,
0.028577432,
-0.0252848044,
0.0050872313,
0.00... |
http://arxiv.org/abs/2407.02301v1 | http://arxiv.org/pdf/2407.02301v1 | CFinBench: A Comprehensive Chinese Financial Benchmark for Large Language Models | 2024-07-02T14:34:36 | 2024-07-02T14:34:36 | Ying Nie, Binwei Yan, Tianyu Guo, Hao Liu, Haoyu Wang, Wei He, Binfan Zheng, Weihao Wang, Qiang Li, Weijian Sun, Yunhe Wang, Dacheng Tao | cs.CL | cs.CL | Large language models (LLMs) have achieved remarkable performance on various
NLP tasks, yet their potential in more challenging and domain-specific task,
such as finance, has not been fully explored. In this paper, we present
CFinBench: a meticulously crafted, the most comprehensive evaluation benchmark
to date, for as... | CFinBench: A Comprehensive Chinese Financial Benchmark for Large Language Models: Large language models (LLMs) have achieved remarkable performance on various
NLP tasks, yet their potential in more challenging and domain-specific task,
such as finance, has not been fully explored. In this paper, we present
CFinBench: ... | [
0.0049450905,
0.0860558674,
-0.0124386344,
0.0064295083,
0.0063922741,
0.0168124828,
0.0294176135,
0.0433872901,
0.0232007429,
-0.0109854992,
-0.0497909375,
-0.0503772795,
0.0063548172,
-0.0022226667,
0.0588468015,
-0.0433281772,
0.0207791235,
0.0040268465,
0.0268474929,
-0.030... |
http://arxiv.org/abs/2407.02284v1 | http://arxiv.org/pdf/2407.02284v1 | Renard: A Modular Pipeline for Extracting Character Networks from Narrative Texts | 2024-07-02T14:14:59 | 2024-07-02T14:14:59 | Arthur Amalvy, Vincent Labatut, Richard Dufour | cs.CL | cs.CL | Renard (Relationships Extraction from NARrative Documents) is a Python
library that allows users to define custom natural language processing (NLP)
pipelines to extract character networks from narrative texts. Contrary to the
few existing tools, Renard can extract dynamic networks, as well as the more
common static net... | Renard: A Modular Pipeline for Extracting Character Networks from Narrative Texts: Renard (Relationships Extraction from NARrative Documents) is a Python
library that allows users to define custom natural language processing (NLP)
pipelines to extract character networks from narrative texts. Contrary to the
few existi... | [
0.0141425729,
0.0570523329,
-0.0380945615,
0.0464136079,
-0.0299644042,
0.0124720708,
0.0079011107,
-0.026585279,
0.0291560926,
-0.0099486411,
0.0573000945,
-0.0147300707,
0.038645409,
0.0423295163,
0.0218183398,
-0.0667713583,
0.0115619255,
0.0000233239,
0.0520983525,
0.042846... |
http://arxiv.org/abs/2407.02273v1 | http://arxiv.org/pdf/2407.02273v1 | Multilingual Trolley Problems for Language Models | 2024-07-02T14:02:53 | 2024-07-02T14:02:53 | Zhijing Jin, Sydney Levine, Max Kleiman-Weiner, Giorgio Piatti, Jiarui Liu, Fernando Gonzalez Adauto, Francesco Ortu, András Strausz, Mrinmaya Sachan, Rada Mihalcea, Yejin Choi, Bernhard Schölkopf | cs.CL | cs.CL | As large language models (LLMs) are deployed in more and more real-world
situations, it is crucial to understand their decision-making when faced with
moral dilemmas. Inspired by a large-scale cross-cultural study of human moral
preferences, "The Moral Machine Experiment", we set up the same set of moral
choices for LL... | Multilingual Trolley Problems for Language Models: As large language models (LLMs) are deployed in more and more real-world
situations, it is crucial to understand their decision-making when faced with
moral dilemmas. Inspired by a large-scale cross-cultural study of human moral
preferences, "The Moral Machine Experime... | [
0.0307175331,
0.1026978567,
-0.034397196,
0.0333824717,
-0.075000219,
0.0322571062,
0.0184855778,
0.0021039511,
-0.0062274653,
0.0048402841,
-0.0062485537,
-0.0206291731,
0.0070480295,
-0.0574978217,
0.0021360288,
-0.0234086812,
0.0237524789,
-0.0213118847,
0.0372461639,
-0.036... |
http://arxiv.org/abs/2407.02243v1 | http://arxiv.org/pdf/2407.02243v1 | Robust Zero-Shot Text-to-Speech Synthesis with Reverse Inference Optimization | 2024-07-02T13:04:04 | 2024-07-02T13:04:04 | Yuchen Hu, Chen Chen, Siyin Wang, Eng Siong Chng, Chao Zhang | cs.CL | cs.CL, cs.SD, eess.AS | In this paper, we propose reverse inference optimization (RIO), a simple and
effective method designed to enhance the robustness of
autoregressive-model-based zero-shot text-to-speech (TTS) systems using
reinforcement learning from human feedback (RLHF). To assess the quality of
speech produced by the TTS system withou... | Robust Zero-Shot Text-to-Speech Synthesis with Reverse Inference Optimization: In this paper, we propose reverse inference optimization (RIO), a simple and
effective method designed to enhance the robustness of
autoregressive-model-based zero-shot text-to-speech (TTS) systems using
reinforcement learning from human fe... | [
-0.0009239232,
0.036295794,
0.0275428351,
-0.0264320374,
-0.013634556,
-0.0245188884,
0.0003577374,
0.017137358,
-0.0375701115,
-0.050887078,
-0.0621120334,
-0.0230025221,
0.0270310417,
-0.0041309223,
0.0417474248,
-0.0764238462,
0.0098177399,
-0.0465254933,
-0.003688676,
-0.00... |
http://arxiv.org/abs/2407.02235v1 | http://arxiv.org/pdf/2407.02235v1 | Towards a Holistic Framework for Multimodal Large Language Models in Three-dimensional Brain CT Report Generation | 2024-07-02T12:58:35 | 2024-07-02T12:58:35 | Cheng-Yi Li, Kao-Jung Chang, Cheng-Fu Yang, Hsin-Yu Wu, Wenting Chen, Hritik Bansal, Ling Chen, Yi-Ping Yang, Yu-Chun Chen, Shih-Pin Chen, Jiing-Feng Lirng, Kai-Wei Chang, Shih-Hwa Chiou | cs.CL | cs.CL | Multi-modal large language models (MLLMs) have been given free rein to
explore exciting medical applications with a primary focus on radiology report
generation. Nevertheless, the preliminary success in 2D radiology captioning is
incompetent to reflect the real-world diagnostic challenge in the volumetric 3D
anatomy. T... | Towards a Holistic Framework for Multimodal Large Language Models in Three-dimensional Brain CT Report Generation: Multi-modal large language models (MLLMs) have been given free rein to
explore exciting medical applications with a primary focus on radiology report
generation. Nevertheless, the preliminary success in 2... | [
0.047739286,
0.0559037253,
0.0263996609,
0.0043456494,
-0.0229227766,
0.0393419117,
0.0158104841,
0.0503829196,
-0.0343888067,
-0.0344330184,
-0.0456733666,
-0.0733755901,
0.0403737165,
0.0599978343,
0.0400160141,
-0.0994001031,
-0.0172848627,
-0.02128876,
-0.0049592578,
0.0041... |
http://arxiv.org/abs/2407.02233v1 | http://arxiv.org/pdf/2407.02233v1 | Synthetic Multimodal Question Generation | 2024-07-02T12:57:42 | 2024-07-02T12:57:42 | Ian Wu, Sravan Jayanthi, Vijay Viswanathan, Simon Rosenberg, Sina Pakazad, Tongshuang Wu, Graham Neubig | cs.CL | cs.CL, cs.AI, cs.LG | Multimodal Retrieval Augmented Generation (MMRAG) is a powerful approach to
question-answering over multimodal documents. A key challenge with evaluating
MMRAG is the paucity of high-quality datasets matching the question styles and
modalities of interest. In light of this, we propose SMMQG, a synthetic data
generation... | Synthetic Multimodal Question Generation: Multimodal Retrieval Augmented Generation (MMRAG) is a powerful approach to
question-answering over multimodal documents. A key challenge with evaluating
MMRAG is the paucity of high-quality datasets matching the question styles and
modalities of interest. In light of this, we ... | [
0.0566712618,
0.0602384843,
-0.0094790282,
0.0262686964,
-0.0161604807,
-0.0084488187,
0.0277106278,
-0.0002318519,
-0.0007778174,
-0.0620234199,
-0.0327302366,
-0.0551910326,
-0.0287251659,
0.0314392075,
0.0670393109,
-0.0331032015,
-0.0095466189,
-0.0138272876,
-0.0537740216,
... |
http://arxiv.org/abs/2407.02211v1 | http://arxiv.org/pdf/2407.02211v1 | PromptIntern: Saving Inference Costs by Internalizing Recurrent Prompt during Large Language Model Fine-tuning | 2024-07-02T12:21:14 | 2024-07-02T12:21:14 | Jiaru Zou, Mengyu Zhou, Tao Li, Shi Han, Dongmei Zhang | cs.CL | cs.CL, cs.AI, cs.LG | Large language models (LLMs) have played a fundamental role in various
natural language processing tasks with powerful prompt techniques. However, in
real-world applications, there are often similar prompt components for repeated
queries, which causes significant computational burdens during inference.
Existing prompt ... | PromptIntern: Saving Inference Costs by Internalizing Recurrent Prompt during Large Language Model Fine-tuning: Large language models (LLMs) have played a fundamental role in various
natural language processing tasks with powerful prompt techniques. However, in
real-world applications, there are often similar prompt c... | [
0.0529499836,
0.0707781836,
-0.0114656975,
0.033041846,
-0.0498896763,
0.0323399901,
-0.0069008768,
0.0024742167,
0.0050451984,
-0.0414248593,
-0.0132168233,
-0.0046237451,
0.0010872401,
0.0179660171,
0.0087949829,
-0.027843453,
0.054590378,
-0.0503397323,
0.0094607212,
-0.0013... |
http://arxiv.org/abs/2407.02209v1 | http://arxiv.org/pdf/2407.02209v1 | Generative Monoculture in Large Language Models | 2024-07-02T12:17:07 | 2024-07-02T12:17:07 | Fan Wu, Emily Black, Varun Chandrasekaran | cs.CL | cs.CL, cs.AI | We introduce {\em generative monoculture}, a behavior observed in large
language models (LLMs) characterized by a significant narrowing of model output
diversity relative to available training data for a given task: for example,
generating only positive book reviews for books with a mixed reception. While
in some cases... | Generative Monoculture in Large Language Models: We introduce {\em generative monoculture}, a behavior observed in large
language models (LLMs) characterized by a significant narrowing of model output
diversity relative to available training data for a given task: for example,
generating only positive book reviews for ... | [
0.0346037149,
0.063853927,
-0.028463291,
0.0215790011,
-0.0368268788,
-0.0031781406,
0.0301166642,
-0.0153467497,
-0.0058732717,
-0.0366394259,
-0.0385286063,
-0.0548981242,
0.0035069908,
0.053124778,
0.0362656824,
-0.0261941198,
0.0571197569,
-0.0344192497,
-0.0221785679,
-0.0... |
http://arxiv.org/abs/2407.02208v1 | http://arxiv.org/pdf/2407.02208v1 | How to Learn in a Noisy World? Self-Correcting the Real-World Data Noise on Machine Translation | 2024-07-02T12:15:15 | 2024-07-02T12:15:15 | Yan Meng, Di Wu, Christof Monz | cs.CL | cs.CL, cs.AI | The massive amounts of web-mined parallel data contain large amounts of
noise. Semantic misalignment, as the primary source of the noise, poses a
challenge for training machine translation systems. In this paper, we first
study the impact of real-world hard-to-detect misalignment noise by proposing a
process to simulat... | How to Learn in a Noisy World? Self-Correcting the Real-World Data Noise on Machine Translation: The massive amounts of web-mined parallel data contain large amounts of
noise. Semantic misalignment, as the primary source of the noise, poses a
challenge for training machine translation systems. In this paper, we first
... | [
0.0325959325,
0.0365075544,
-0.0023592049,
0.0323412903,
-0.0021333543,
0.0281859152,
0.0064005717,
-0.009576818,
-0.0339701287,
-0.0359276831,
-0.047691308,
-0.0291491635,
0.0353235118,
-0.046950642,
0.0357748792,
-0.0249104165,
0.0398392268,
-0.0143818045,
0.0027704807,
-0.01... |
http://arxiv.org/abs/2407.02203v1 | http://arxiv.org/pdf/2407.02203v1 | Automatic Adaptation Rule Optimization via Large Language Models | 2024-07-02T12:06:40 | 2024-07-02T12:06:40 | Yusei Ishimizu, Jialong Li, Jinglue Xu, Jinyu Cai, Hitoshi Iba, Kenji Tei | cs.CL | cs.CL, cs.AI | Rule-based adaptation is a foundational approach to self-adaptation,
characterized by its human readability and rapid response. However, building
high-performance and robust adaptation rules is often a challenge because it
essentially involves searching the optimal design in a complex (variables)
space. In response, th... | Automatic Adaptation Rule Optimization via Large Language Models: Rule-based adaptation is a foundational approach to self-adaptation,
characterized by its human readability and rapid response. However, building
high-performance and robust adaptation rules is often a challenge because it
essentially involves searching ... | [
0.0198695268,
0.0060729524,
-0.0456816666,
0.0079759052,
0.0147062652,
0.0019364286,
-0.0016453834,
0.0458840579,
-0.0079648336,
-0.0501806028,
-0.014655238,
-0.0101731429,
-0.0029852374,
0.0260281023,
0.0161827132,
-0.0311660003,
0.0277436543,
-0.0749620646,
0.0126276156,
0.00... |
http://arxiv.org/abs/2407.02147v1 | http://arxiv.org/pdf/2407.02147v1 | LlamAr & GemmAr: Enhancing LLMs Through Arabic Instruction-Tuning | 2024-07-02T10:43:49 | 2024-07-02T10:43:49 | Hasna Chouikhi, Manel Aloui, Cyrine Ben Hammou, Ghaith Chaabane, Haithem Kchaou, Chehir Dhaouadi | cs.CL | cs.CL, cs.AI | Large language models (LLMs) have greatly impacted the natural language
processing (NLP) field, particularly for the English language. These models
have demonstrated capabilities in understanding and generating human-like text.
The success of language models largely depends on the availability of
high-quality instructi... | LlamAr & GemmAr: Enhancing LLMs Through Arabic Instruction-Tuning: Large language models (LLMs) have greatly impacted the natural language
processing (NLP) field, particularly for the English language. These models
have demonstrated capabilities in understanding and generating human-like text.
The success of language m... | [
0.0494179949,
0.0424378328,
-0.0059583802,
0.0125336219,
-0.0504959226,
0.024580529,
0.0450680256,
-0.0394820832,
-0.0194679499,
-0.058244098,
-0.044025775,
-0.0786099136,
0.0334228724,
0.0060161385,
0.0342155285,
-0.0395956337,
0.0059015318,
-0.0807836354,
-0.0198006798,
0.010... |
http://arxiv.org/abs/2407.02138v1 | http://arxiv.org/pdf/2407.02138v1 | Efficient Nearest Neighbor based Uncertainty Estimation for Natural Language Processing Tasks | 2024-07-02T10:33:31 | 2024-07-02T10:33:31 | Wataru Hashimoto, Hidetaka Kamigaito, Taro Watanabe | cs.CL | cs.CL, cs.AI, cs.LG | Trustworthy prediction in Deep Neural Networks (DNNs), including Pre-trained
Language Models (PLMs) is important for safety-critical applications in the
real world. However, DNNs often suffer from uncertainty estimation, such as
miscalibration. In particular, approaches that require multiple stochastic
inference can mi... | Efficient Nearest Neighbor based Uncertainty Estimation for Natural Language Processing Tasks: Trustworthy prediction in Deep Neural Networks (DNNs), including Pre-trained
Language Models (PLMs) is important for safety-critical applications in the
real world. However, DNNs often suffer from uncertainty estimation, suc... | [
0.0417193025,
0.0946857706,
0.0273566134,
0.0590956174,
-0.0206637271,
0.0021073816,
0.0203617234,
-0.0056573297,
0.0006847941,
-0.0017712043,
-0.0522074215,
-0.0779992938,
0.0099538453,
-0.007442012,
0.0050319876,
-0.0049435361,
0.0031436703,
-0.0276173316,
0.0353059694,
-0.03... |
http://arxiv.org/abs/2407.02136v1 | http://arxiv.org/pdf/2407.02136v1 | Black Big Boxes: Do Language Models Hide a Theory of Adjective Order? | 2024-07-02T10:29:09 | 2024-07-02T10:29:09 | Jaap Jumelet, Lisa Bylinina, Willem Zuidema, Jakub Szymanik | cs.CL | cs.CL | In English and other languages, multiple adjectives in a complex noun phrase
show intricate ordering patterns that have been a target of much linguistic
theory. These patterns offer an opportunity to assess the ability of language
models (LMs) to learn subtle rules of language involving factors that cross the
tradition... | Black Big Boxes: Do Language Models Hide a Theory of Adjective Order?: In English and other languages, multiple adjectives in a complex noun phrase
show intricate ordering patterns that have been a target of much linguistic
theory. These patterns offer an opportunity to assess the ability of language
models (LMs) to le... | [
0.0663952753,
0.0844389796,
-0.0476068407,
0.0406468399,
-0.0209996607,
0.0351970159,
-0.0417758115,
0.0033966159,
-0.0318369865,
-0.061400272,
-0.0174111985,
-0.0640671253,
-0.0020300981,
-0.0211417973,
0.0170013364,
-0.0309446696,
0.0011174661,
-0.0243201647,
-0.012625603,
-0... |
http://arxiv.org/abs/2407.02122v1 | http://arxiv.org/pdf/2407.02122v1 | Fake News Detection: It's All in the Data! | 2024-07-02T10:12:06 | 2024-07-02T10:12:06 | Soveatin Kuntur, Anna Wróblewska, Marcin Paprzycki, Maria Ganzha | cs.CL | cs.CL | This comprehensive survey serves as an indispensable resource for researchers
embarking on the journey of fake news detection. By highlighting the pivotal
role of dataset quality and diversity, it underscores the significance of these
elements in the effectiveness and robustness of detection models. The survey
meticulo... | Fake News Detection: It's All in the Data!: This comprehensive survey serves as an indispensable resource for researchers
embarking on the journey of fake news detection. By highlighting the pivotal
role of dataset quality and diversity, it underscores the significance of these
elements in the effectiveness and robustn... | [
0.0777116418,
0.0347211212,
-0.027682256,
0.0529901981,
-0.0402635373,
0.0210758895,
0.0164639968,
0.00892526,
0.0211547669,
-0.0034068164,
-0.0010892201,
-0.0137260985,
0.0042376928,
0.0762140825,
-0.0037060482,
-0.0047678393,
0.0583684184,
0.0413318425,
0.0614006408,
-0.01112... |
http://arxiv.org/abs/2407.02119v1 | http://arxiv.org/pdf/2407.02119v1 | Cost-Effective Proxy Reward Model Construction with On-Policy and Active Learning | 2024-07-02T10:09:19 | 2024-07-02T10:09:19 | Yifang Chen, Shuohang Wang, Ziyi Yang, Hiteshi Sharma, Nikos Karampatziakis, Donghan Yu, Kevin Jamieson, Simon Shaolei Du, Yelong Shen | cs.LG | cs.LG, cs.AI, cs.CL | Reinforcement learning with human feedback (RLHF), as a widely adopted
approach in current large language model pipelines, is \textit{bottlenecked by
the size of human preference data}. While traditional methods rely on offline
preference dataset constructions, recent approaches have shifted towards online
settings, wh... | Cost-Effective Proxy Reward Model Construction with On-Policy and Active Learning: Reinforcement learning with human feedback (RLHF), as a widely adopted
approach in current large language model pipelines, is \textit{bottlenecked by
the size of human preference data}. While traditional methods rely on offline
preferen... | [
0.0630529895,
0.0668731481,
-0.0057568573,
-0.0274291988,
-0.0090795876,
0.0221100319,
-0.0016030137,
-0.0089955004,
-0.0088691609,
-0.0295946691,
-0.0662756786,
-0.0305312015,
-0.0204138365,
0.0259758178,
0.0078741685,
-0.069333002,
0.0421967544,
-0.0490449257,
0.0526687168,
-... |
http://arxiv.org/abs/2407.02118v1 | http://arxiv.org/pdf/2407.02118v1 | Breaking Language Barriers: Cross-Lingual Continual Pre-Training at Scale | 2024-07-02T10:06:41 | 2024-07-02T10:06:41 | Wenzhen Zheng, Wenbo Pan, Xu Xu, Libo Qin, Li Yue, Ming Zhou | cs.CL | cs.CL | In recent years, Large Language Models (LLMs) have made significant strides
towards Artificial General Intelligence. However, training these models from
scratch requires substantial computational resources and vast amounts of text
data. In this paper, we explore an alternative approach to constructing an LLM
for a new ... | Breaking Language Barriers: Cross-Lingual Continual Pre-Training at Scale: In recent years, Large Language Models (LLMs) have made significant strides
towards Artificial General Intelligence. However, training these models from
scratch requires substantial computational resources and vast amounts of text
data. In this... | [
0.0012048916,
0.0512964949,
-0.0191119276,
0.0130652804,
-0.0246450119,
0.02812621,
0.0190564096,
-0.0244461019,
-0.0037925444,
-0.0202582106,
-0.0428609848,
-0.0834076405,
-0.0243611969,
0.0348800234,
-0.0027045766,
-0.0399433672,
0.0496180989,
-0.0580642559,
-0.0339321382,
-0... |
http://arxiv.org/abs/2407.02099v1 | http://arxiv.org/pdf/2407.02099v1 | Helpful assistant or fruitful facilitator? Investigating how personas affect language model behavior | 2024-07-02T09:36:54 | 2024-07-02T09:36:54 | Pedro Henrique Luz de Araujo, Benjamin Roth | cs.CL | cs.CL | One way to personalize and steer generations from large language models (LLM)
is to assign a persona: a role that describes how the user expects the LLM to
behave (e.g., a helpful assistant, a teacher, a woman). This paper investigates
how personas affect diverse aspects of model behavior. We assign to seven LLMs
162 p... | Helpful assistant or fruitful facilitator? Investigating how personas affect language model behavior: One way to personalize and steer generations from large language models (LLM)
is to assign a persona: a role that describes how the user expects the LLM to
behave (e.g., a helpful assistant, a teacher, a woman). This ... | [
0.0887645185,
0.0489915609,
-0.0474096611,
0.0122965192,
-0.0120983981,
0.0005452876,
0.0456561558,
-0.0228090268,
-0.0028106393,
-0.0133944266,
0.0144649968,
-0.057062,
-0.0312040541,
-0.0087631196,
0.0183596686,
-0.0702520385,
0.0444598794,
-0.0093445368,
0.0508798435,
-0.011... |
http://arxiv.org/abs/2407.02067v1 | http://arxiv.org/pdf/2407.02067v1 | Crossroads of Continents: Automated Artifact Extraction for Cultural Adaptation with Large Multimodal Models | 2024-07-02T08:55:41 | 2024-07-02T08:55:41 | Anjishnu Mukherjee, Ziwei Zhu, Antonios Anastasopoulos | cs.CL | cs.CL | In this work, we present a comprehensive three-phase study to examine (1) the
effectiveness of large multimodal models (LMMs) in recognizing cultural
contexts; (2) the accuracy of their representations of diverse cultures; and
(3) their ability to adapt content across cultural boundaries. We first
introduce Dalle Stree... | Crossroads of Continents: Automated Artifact Extraction for Cultural Adaptation with Large Multimodal Models: In this work, we present a comprehensive three-phase study to examine (1) the
effectiveness of large multimodal models (LMMs) in recognizing cultural
contexts; (2) the accuracy of their representations of dive... | [
0.0132092265,
0.0690411106,
-0.0280508045,
0.0563852303,
-0.0193523914,
0.0535215251,
0.0458396189,
-0.0326444171,
0.0137244482,
-0.0411962084,
-0.0175740663,
-0.0044040806,
-0.0127794314,
0.0087020965,
0.0174770299,
-0.1155683622,
-0.0027371077,
-0.0180624761,
-0.0491048694,
0... |
http://arxiv.org/abs/2407.02066v1 | http://arxiv.org/pdf/2407.02066v1 | BiasDora: Exploring Hidden Biased Associations in Vision-Language Models | 2024-07-02T08:55:40 | 2024-07-02T08:55:40 | Chahat Raj, Anjishnu Mukherjee, Aylin Caliskan, Antonios Anastasopoulos, Ziwei Zhu | cs.CL | cs.CL | Existing works examining Vision Language Models (VLMs) for social biases
predominantly focus on a limited set of documented bias associations, such as
gender:profession or race:crime. This narrow scope often overlooks a vast range
of unexamined implicit associations, restricting the identification and, hence,
mitigatio... | BiasDora: Exploring Hidden Biased Associations in Vision-Language Models: Existing works examining Vision Language Models (VLMs) for social biases
predominantly focus on a limited set of documented bias associations, such as
gender:profession or race:crime. This narrow scope often overlooks a vast range
of unexamined i... | [
0.016458096,
0.0863892063,
-0.040753141,
0.051319927,
0.0017550914,
0.0384856872,
0.0463110991,
0.0075518927,
0.0485959835,
-0.0747367442,
-0.0353193432,
0.0294606518,
0.0407766774,
0.0412508175,
0.0172843002,
-0.0640371069,
0.0476960875,
0.0289164819,
0.0274459235,
-0.02043295... |
http://arxiv.org/abs/2407.02062v1 | http://arxiv.org/pdf/2407.02062v1 | Are Data Augmentation Methods in Named Entity Recognition Applicable for Uncertainty Estimation? | 2024-07-02T08:49:43 | 2024-07-02T08:49:43 | Wataru Hashimoto, Hidetaka Kamigaito, Taro Watanabe | cs.CL | cs.CL, cs.AI, cs.LG | This work investigates the impact of data augmentation on confidence
calibration and uncertainty estimation in Named Entity Recognition (NER) tasks.
For the future advance of NER in safety-critical fields like healthcare and
finance, it is essential to achieve accurate predictions with calibrated
confidence when applyi... | Are Data Augmentation Methods in Named Entity Recognition Applicable for Uncertainty Estimation?: This work investigates the impact of data augmentation on confidence
calibration and uncertainty estimation in Named Entity Recognition (NER) tasks.
For the future advance of NER in safety-critical fields like healthcare ... | [
0.0106259631,
0.0952430218,
0.0131428381,
0.0272325817,
0.0232018325,
0.0303126033,
0.0120254271,
-0.0305000991,
-0.030024115,
0.0195554998,
-0.0414004326,
-0.0845756084,
0.0247803032,
0.0267427638,
0.0501229092,
-0.0249357801,
0.02896888,
-0.0181777626,
0.0547740422,
0.0157920... |
http://arxiv.org/abs/2407.02056v1 | http://arxiv.org/pdf/2407.02056v1 | Integrate the Essence and Eliminate the Dross: Fine-Grained Self-Consistency for Free-Form Language Generation | 2024-07-02T08:38:31 | 2024-07-02T08:38:31 | Xinglin Wang, Yiwei Li, Shaoxiong Feng, Peiwen Yuan, Boyuan Pan, Heda Wang, Yao Hu, Kan Li | cs.CL | cs.CL, cs.AI | Self-consistency (SC), leveraging multiple samples from LLMs, shows
significant gains on various reasoning tasks but struggles with free-form
generation due to the difficulty of aggregating answers. Its variants, UCS and
USC, rely on sample selection or voting mechanisms to improve output quality.
These methods, howeve... | Integrate the Essence and Eliminate the Dross: Fine-Grained Self-Consistency for Free-Form Language Generation: Self-consistency (SC), leveraging multiple samples from LLMs, shows
significant gains on various reasoning tasks but struggles with free-form
generation due to the difficulty of aggregating answers. Its vari... | [
-0.0210756212,
0.0478501655,
0.0074160667,
0.0032862644,
-0.0372137763,
-0.0145694278,
-0.0293269716,
-0.0192602035,
-0.0204537306,
-0.0191090256,
-0.0198693816,
0.0107598798,
0.017393101,
0.0120559214,
0.03712276,
-0.0178832319,
0.0330319889,
-0.0202446114,
-0.03951535,
-0.028... |
http://arxiv.org/abs/2407.02049v1 | http://arxiv.org/pdf/2407.02049v1 | Accompanied Singing Voice Synthesis with Fully Text-controlled Melody | 2024-07-02T08:23:38 | 2024-07-02T08:23:38 | Ruiqi Li, Zhiqing Hong, Yongqi Wang, Lichao Zhang, Rongjie Huang, Siqi Zheng, Zhou Zhao | eess.AS | eess.AS, cs.CL, cs.SD | Text-to-song (TTSong) is a music generation task that synthesizes accompanied
singing voices. Current TTSong methods, inherited from singing voice synthesis
(SVS), require melody-related information that can sometimes be impractical,
such as music scores or MIDI sequences. We present MelodyLM, the first TTSong
model th... | Accompanied Singing Voice Synthesis with Fully Text-controlled Melody: Text-to-song (TTSong) is a music generation task that synthesizes accompanied
singing voices. Current TTSong methods, inherited from singing voice synthesis
(SVS), require melody-related information that can sometimes be impractical,
such as music s... | [
0.0068906113,
0.0157572906,
0.0120226461,
-0.0331445858,
-0.0258180853,
-0.0177116059,
0.0505386405,
-0.0260042492,
-0.0423109904,
0.0404658653,
-0.0581202246,
-0.042259559,
0.0029001683,
0.0248160642,
0.0426185802,
-0.0701787919,
0.0064649084,
-0.0308671575,
0.0114743412,
-0.0... |
http://arxiv.org/abs/2407.02043v1 | http://arxiv.org/pdf/2407.02043v1 | Concise and Precise Context Compression for Tool-Using Language Models | 2024-07-02T08:17:00 | 2024-07-02T08:17:00 | Yang Xu, Yunlong Feng, Honglin Mu, Yutai Hou, Yitong Li, Xinghao Wang, Wanjun Zhong, Zhongyang Li, Dandan Tu, Qingfu Zhu, Min Zhang, Wanxiang Che | cs.CL | cs.CL | Through reading the documentation in the context, tool-using language models
can dynamically extend their capability using external tools. The cost is that
we have to input lengthy documentation every time the model needs to use the
tool, occupying the input window as well as slowing down the decoding process.
Given ... | Concise and Precise Context Compression for Tool-Using Language Models: Through reading the documentation in the context, tool-using language models
can dynamically extend their capability using external tools. The cost is that
we have to input lengthy documentation every time the model needs to use the
tool, occupying... | [
-0.0059703737,
0.023948295,
-0.0270769149,
0.0040204334,
-0.0306782965,
0.002010979,
-0.040015392,
0.0552699976,
-0.1044513732,
-0.0120199081,
0.018997075,
0.0348086767,
0.0240920167,
-0.0034819944,
0.0275891218,
-0.0616946518,
0.030709181,
-0.0006338144,
-0.0103985108,
0.03274... |
http://arxiv.org/abs/2407.02042v1 | http://arxiv.org/pdf/2407.02042v1 | Fake News Detection and Manipulation Reasoning via Large Vision-Language Models | 2024-07-02T08:16:43 | 2024-07-02T08:16:43 | Ruihan Jin, Ruibo Fu, Zhengqi Wen, Shuai Zhang, Yukun Liu, Jianhua Tao | cs.CL | cs.CL, cs.AI | Fake news becomes a growing threat to information security and public opinion
with the rapid sprawl of media manipulation. Therefore, fake news detection
attracts widespread attention from academic community. Traditional fake news
detection models demonstrate remarkable performance on authenticity binary
classification... | Fake News Detection and Manipulation Reasoning via Large Vision-Language Models: Fake news becomes a growing threat to information security and public opinion
with the rapid sprawl of media manipulation. Therefore, fake news detection
attracts widespread attention from academic community. Traditional fake news
detecti... | [
0.0651042387,
0.0093470355,
0.017312618,
0.0682549477,
-0.0555309765,
0.0258184541,
-0.003012982,
0.0013379981,
0.0040837736,
-0.0646873638,
-0.0055706045,
-0.0193044692,
0.0194084104,
0.0633223653,
0.0444770195,
-0.0394084007,
0.0758899525,
0.0201468207,
0.0423679911,
0.012174... |
http://arxiv.org/abs/2407.02039v1 | http://arxiv.org/pdf/2407.02039v1 | Prompt Stability Scoring for Text Annotation with Large Language Models | 2024-07-02T08:11:18 | 2024-07-02T08:11:18 | Christopher Barrie, Elli Palaiologou, Petter Törnberg | cs.CL | cs.CL | Researchers are increasingly using language models (LMs) for text annotation.
These approaches rely only on a prompt telling the model to return a given
output according to a set of instructions. The reproducibility of LM outputs
may nonetheless be vulnerable to small changes in the prompt design. This calls
into quest... | Prompt Stability Scoring for Text Annotation with Large Language Models: Researchers are increasingly using language models (LMs) for text annotation.
These approaches rely only on a prompt telling the model to return a given
output according to a set of instructions. The reproducibility of LM outputs
may nonetheless b... | [
0.0321040042,
0.0112718157,
-0.0038922112,
0.0039386707,
-0.0072265463,
0.0319598578,
0.0222303905,
0.0056263418,
0.0331612118,
-0.0524133071,
0.0051260171,
-0.0542583242,
0.0312247518,
0.0279508773,
0.0099196695,
-0.0380078033,
0.0581143014,
-0.0488309897,
0.0113177411,
0.0305... |
http://arxiv.org/abs/2407.02030v1 | http://arxiv.org/pdf/2407.02030v1 | Breaking Bias, Building Bridges: Evaluation and Mitigation of Social Biases in LLMs via Contact Hypothesis | 2024-07-02T07:58:46 | 2024-07-02T07:58:46 | Chahat Raj, Anjishnu Mukherjee, Aylin Caliskan, Antonios Anastasopoulos, Ziwei Zhu | cs.CL | cs.CL | Large Language Models (LLMs) perpetuate social biases, reflecting prejudices
in their training data and reinforcing societal stereotypes and inequalities.
Our work explores the potential of the Contact Hypothesis, a concept from
social psychology for debiasing LLMs. We simulate various forms of social
contact through L... | Breaking Bias, Building Bridges: Evaluation and Mitigation of Social Biases in LLMs via Contact Hypothesis: Large Language Models (LLMs) perpetuate social biases, reflecting prejudices
in their training data and reinforcing societal stereotypes and inequalities.
Our work explores the potential of the Contact Hypothesi... | [
0.0391403399,
0.0504559651,
-0.0426324457,
0.0526030697,
0.0102584455,
0.029253222,
0.0431392118,
-0.0173512381,
0.0629565194,
-0.0356683917,
-0.0629229248,
-0.037117891,
0.0388337597,
-0.0117995311,
0.0536659583,
-0.0497085154,
0.0644874945,
0.0004347672,
0.0439061336,
-0.0509... |
http://arxiv.org/abs/2407.02028v1 | http://arxiv.org/pdf/2407.02028v1 | Why does in-context learning fail sometimes? Evaluating in-context learning on open and closed questions | 2024-07-02T07:52:30 | 2024-07-02T07:52:30 | Xiang Li, Haoran Tang, Siyu Chen, Ziwei Wang, Ryan Chen, Marcin Abram | cs.CL | cs.CL, cs.AI, cs.IR, cs.LG | We measure the performance of in-context learning as a function of task
novelty and difficulty for open and closed questions. For that purpose, we
created a novel benchmark consisting of hard scientific questions, each paired
with a context of various relevancy. We show that counter-intuitively, a
context that is more ... | Why does in-context learning fail sometimes? Evaluating in-context learning on open and closed questions: We measure the performance of in-context learning as a function of task
novelty and difficulty for open and closed questions. For that purpose, we
created a novel benchmark consisting of hard scientific questions,... | [
0.01724343,
0.0116413366,
0.0001442925,
0.0242623463,
-0.0281813946,
0.0255072024,
0.0239167679,
-0.0074558994,
-0.0197242834,
-0.0667991415,
-0.0133324964,
-0.0406335965,
-0.0379908197,
0.0036162338,
0.0276375711,
-0.0181725211,
0.0356218368,
-0.0131056178,
-0.0095080854,
0.00... |
http://arxiv.org/abs/2407.02005v1 | http://arxiv.org/pdf/2407.02005v1 | An End-to-End Speech Summarization Using Large Language Model | 2024-07-02T07:22:57 | 2024-07-02T07:22:57 | Hengchao Shang, Zongyao Li, Jiaxin Guo, Shaojun Li, Zhiqiang Rao, Yuanchang Luo, Daimeng Wei, Hao Yang | cs.CL | cs.CL, cs.SD, eess.AS | Abstractive Speech Summarization (SSum) aims to generate human-like text
summaries from spoken content. It encounters difficulties in handling long
speech input and capturing the intricate cross-modal mapping between long
speech inputs and short text summaries. Research on large language models
(LLMs) and multimodal in... | An End-to-End Speech Summarization Using Large Language Model: Abstractive Speech Summarization (SSum) aims to generate human-like text
summaries from spoken content. It encounters difficulties in handling long
speech input and capturing the intricate cross-modal mapping between long
speech inputs and short text summar... | [
0.0187649857,
0.0058853561,
0.0398490988,
0.0098644923,
0.0092539024,
-0.0057548368,
-0.0275075752,
-0.0445849225,
-0.034513142,
-0.0597742274,
-0.0479716957,
-0.0298655834,
0.0477000214,
0.0031692409,
0.0282026455,
-0.090426296,
-0.0052452623,
-0.0587203689,
-0.0490288325,
-0.... |
http://arxiv.org/abs/2407.02543v1 | http://arxiv.org/pdf/2407.02543v1 | Towards the Next Frontier in Speech Representation Learning Using Disentanglement | 2024-07-02T07:13:35 | 2024-07-02T07:13:35 | Varun Krishna, Sriram Ganapathy | cs.CL | cs.CL, cs.AI, cs.LG, cs.SD, eess.AS | The popular frameworks for self-supervised learning of speech representations
have largely focused on frame-level masked prediction of speech regions. While
this has shown promising downstream task performance for speech recognition and
related tasks, this has largely ignored factors of speech that are encoded at
coars... | Towards the Next Frontier in Speech Representation Learning Using Disentanglement: The popular frameworks for self-supervised learning of speech representations
have largely focused on frame-level masked prediction of speech regions. While
this has shown promising downstream task performance for speech recognition and... | [
-0.0524283461,
0.0704809353,
-0.0033737218,
0.0348762833,
-0.0036785735,
0.0101905344,
0.0261714365,
-0.0166521855,
0.0336993448,
-0.0140157882,
-0.0778663903,
0.0074167559,
0.0093915276,
0.0311593805,
0.0049806107,
-0.0504128411,
0.0344330147,
-0.0113154799,
-0.0428627133,
-0.... |
http://arxiv.org/abs/2407.01994v1 | http://arxiv.org/pdf/2407.01994v1 | Simple Augmentations of Logical Rules for Neuro-Symbolic Knowledge Graph Completion | 2024-07-02T07:07:59 | 2024-07-02T07:07:59 | Ananjan Nandi, Navdeep Kaur, Parag Singla, Mausam | cs.AI | cs.AI, cs.CL, cs.IR | High-quality and high-coverage rule sets are imperative to the success of
Neuro-Symbolic Knowledge Graph Completion (NS-KGC) models, because they form
the basis of all symbolic inferences. Recent literature builds neural models
for generating rule sets, however, preliminary experiments show that they
struggle with main... | Simple Augmentations of Logical Rules for Neuro-Symbolic Knowledge Graph Completion: High-quality and high-coverage rule sets are imperative to the success of
Neuro-Symbolic Knowledge Graph Completion (NS-KGC) models, because they form
the basis of all symbolic inferences. Recent literature builds neural models
for ge... | [
0.0555141792,
0.0298944823,
-0.033444453,
0.020172894,
-0.0449275374,
-0.0087482575,
-0.049828209,
0.0241865329,
0.0343564115,
0.0359603316,
0.0669231638,
0.0427393056,
-0.0047427421,
0.0363726951,
0.0340335853,
-0.0040863897,
0.0493996032,
-0.0048768283,
0.0325477086,
-0.01750... |
http://arxiv.org/abs/2407.01992v1 | http://arxiv.org/pdf/2407.01992v1 | Is Your Large Language Model Knowledgeable or a Choices-Only Cheater? | 2024-07-02T07:06:53 | 2024-07-02T07:06:53 | Nishant Balepur, Rachel Rudinger | cs.CL | cs.CL | Recent work shows that large language models (LLMs) can answer
multiple-choice questions using only the choices, but does this mean that MCQA
leaderboard rankings of LLMs are largely influenced by abilities in
choices-only settings? To answer this, we use a contrast set that probes if
LLMs over-rely on choices-only sho... | Is Your Large Language Model Knowledgeable or a Choices-Only Cheater?: Recent work shows that large language models (LLMs) can answer
multiple-choice questions using only the choices, but does this mean that MCQA
leaderboard rankings of LLMs are largely influenced by abilities in
choices-only settings? To answer this, ... | [
0.014815595,
0.0569116622,
-0.0217824541,
0.038429141,
-0.0311317425,
0.0120606786,
-0.0145750558,
0.001654075,
0.0578058623,
-0.032307595,
-0.0270014871,
0.0377001315,
-0.0075152507,
-0.014455351,
0.0441082381,
-0.0044309702,
0.0128175011,
-0.0538671426,
-0.0133769652,
-0.0311... |
http://arxiv.org/abs/2407.01976v1 | http://arxiv.org/pdf/2407.01976v1 | A Bounding Box is Worth One Token: Interleaving Layout and Text in a Large Language Model for Document Understanding | 2024-07-02T06:29:05 | 2024-07-02T06:29:05 | Jinghui Lu, Haiyang Yu, Yanjie Wang, Yongjie Ye, Jingqun Tang, Ziwei Yang, Binghong Wu, Qi Liu, Hao Feng, Han Wang, Hao Liu, Can Huang | cs.CL | cs.CL, cs.AI, cs.MM | Recently, many studies have demonstrated that exclusively incorporating
OCR-derived text and spatial layouts with large language models (LLMs) can be
highly effective for document understanding tasks. However, existing methods
that integrate spatial layouts with text have limitations, such as producing
overly long text... | A Bounding Box is Worth One Token: Interleaving Layout and Text in a Large Language Model for Document Understanding: Recently, many studies have demonstrated that exclusively incorporating
OCR-derived text and spatial layouts with large language models (LLMs) can be
highly effective for document understanding tasks. ... | [
0.0169268735,
0.0124651678,
0.0333785303,
0.0655791238,
-0.0265173614,
0.0346932709,
0.0263839774,
0.0029711551,
0.0323761255,
-0.0348234214,
-0.0252454076,
-0.0053320639,
0.0179548617,
0.0680654123,
-0.0016830933,
-0.0430300161,
0.0189273972,
-0.0782426596,
-0.0018269182,
0.01... |
http://arxiv.org/abs/2407.01965v1 | http://arxiv.org/pdf/2407.01965v1 | AdaCQR: Enhancing Query Reformulation for Conversational Search via Sparse and Dense Retrieval Alignment | 2024-07-02T05:50:16 | 2024-07-02T05:50:16 | Yilong Lai, Jialong Wu, Congzhi Zhang, Haowen Sun, Deyu Zhou | cs.CL | cs.CL, cs.IR | Conversational Query Reformulation (CQR) has significantly advanced in
addressing the challenges of conversational search, particularly those stemming
from the latent user intent and the need for historical context. Recent works
aimed to boost the performance of CRQ through alignment. However, they are
designed for one... | AdaCQR: Enhancing Query Reformulation for Conversational Search via Sparse and Dense Retrieval Alignment: Conversational Query Reformulation (CQR) has significantly advanced in
addressing the challenges of conversational search, particularly those stemming
from the latent user intent and the need for historical contex... | [
0.0602645464,
0.058520589,
0.000206769,
-0.0061052423,
0.0056901025,
-0.0252160914,
-0.0266301837,
0.019455051,
-0.0340745561,
-0.0275153369,
-0.0590605773,
-0.0430897549,
-0.0362620205,
0.0241697337,
0.002289081,
-0.0399931855,
0.021956766,
0.0292961188,
-0.0128716687,
0.05396... |
http://arxiv.org/abs/2407.01964v2 | http://arxiv.org/pdf/2407.01964v2 | Enabling Discriminative Reasoning in LLMs for Legal Judgment Prediction | 2024-07-02T05:43:15 | 2024-07-03T02:25:23 | Chenlong Deng, Kelong Mao, Yuyao Zhang, Zhicheng Dou | cs.CL | cs.CL | Legal judgment prediction is essential for enhancing judicial efficiency. In
this work, we identify that existing large language models (LLMs) underperform
in this domain due to challenges in understanding case complexities and
distinguishing between similar charges. To adapt LLMs for effective legal
judgment predictio... | Enabling Discriminative Reasoning in LLMs for Legal Judgment Prediction: Legal judgment prediction is essential for enhancing judicial efficiency. In
this work, we identify that existing large language models (LLMs) underperform
in this domain due to challenges in understanding case complexities and
distinguishing betw... | [
0.0623039752,
0.0395288691,
-0.0137967477,
0.0375402644,
-0.0521018095,
0.0196380094,
0.0191434864,
0.0306138694,
0.0241115969,
-0.0830811709,
-0.0370315537,
-0.0551199131,
-0.0370267518,
-0.0109718051,
-0.0128628947,
-0.0483716168,
0.0562874712,
-0.0238462333,
0.1033281162,
0.... |
http://arxiv.org/abs/2407.01955v1 | http://arxiv.org/pdf/2407.01955v1 | S2D: Sorted Speculative Decoding For More Efficient Deployment of Nested Large Language Models | 2024-07-02T05:14:15 | 2024-07-02T05:14:15 | Parsa Kavehzadeh, Mohammadreza Pourreza, Mojtaba Valipour, Tinashu Zhu, Haoli Bai, Ali Ghodsi, Boxing Chen, Mehdi Rezagholizadeh | cs.CL | cs.CL | Deployment of autoregressive large language models (LLMs) is costly, and as
these models increase in size, the associated costs will become even more
considerable. Consequently, different methods have been proposed to accelerate
the token generation process and reduce costs. Speculative decoding (SD) is
among the most ... | S2D: Sorted Speculative Decoding For More Efficient Deployment of Nested Large Language Models: Deployment of autoregressive large language models (LLMs) is costly, and as
these models increase in size, the associated costs will become even more
considerable. Consequently, different methods have been proposed to accel... | [
0.0227438603,
0.096274741,
-0.000669048,
0.0258988738,
-0.0267639998,
0.013517851,
0.0162890665,
0.0121107856,
-0.0539210513,
-0.0422587059,
-0.00786395,
-0.0844819918,
0.0049321186,
0.0463211909,
-0.0051653273,
-0.0061893733,
0.0321512297,
-0.0633746982,
0.0162279382,
-0.00785... |
http://arxiv.org/abs/2407.01948v1 | http://arxiv.org/pdf/2407.01948v1 | Extracting and Encoding: Leveraging Large Language Models and Medical Knowledge to Enhance Radiological Text Representation | 2024-07-02T04:39:19 | 2024-07-02T04:39:19 | Pablo Messina, René Vidal, Denis Parra, Álvaro Soto, Vladimir Araujo | cs.CL | cs.CL, cs.AI, cs.LG | Advancing representation learning in specialized fields like medicine remains
challenging due to the scarcity of expert annotations for text and images. To
tackle this issue, we present a novel two-stage framework designed to extract
high-quality factual statements from free-text radiology reports in order to
improve t... | Extracting and Encoding: Leveraging Large Language Models and Medical Knowledge to Enhance Radiological Text Representation: Advancing representation learning in specialized fields like medicine remains
challenging due to the scarcity of expert annotations for text and images. To
tackle this issue, we present a novel ... | [
0.0300117005,
0.0485619865,
0.0458548814,
0.0165282674,
-0.0131989475,
0.044652693,
0.0018515879,
0.0399656221,
-0.012211794,
-0.0250686463,
-0.0107089942,
-0.0379071049,
0.0121299792,
0.0254311543,
0.0324934982,
-0.0605179034,
0.0105293328,
0.0209341738,
0.0692063272,
0.004965... |
http://arxiv.org/abs/2407.01942v1 | http://arxiv.org/pdf/2407.01942v1 | Certainly Uncertain: A Benchmark and Metric for Multimodal Epistemic and Aleatoric Awareness | 2024-07-02T04:23:54 | 2024-07-02T04:23:54 | Khyathi Raghavi Chandu, Linjie Li, Anas Awadalla, Ximing Lu, Jae Sung Park, Jack Hessel, Lijuan Wang, Yejin Choi | cs.AI | cs.AI, cs.CL, cs.CV | The ability to acknowledge the inevitable uncertainty in their knowledge and
reasoning is a prerequisite for AI systems to be truly truthful and reliable.
In this paper, we present a taxonomy of uncertainty specific to vision-language
AI systems, distinguishing between epistemic uncertainty (arising from a lack
of info... | Certainly Uncertain: A Benchmark and Metric for Multimodal Epistemic and Aleatoric Awareness: The ability to acknowledge the inevitable uncertainty in their knowledge and
reasoning is a prerequisite for AI systems to be truly truthful and reliable.
In this paper, we present a taxonomy of uncertainty specific to vision... | [
0.0287699867,
0.0885877758,
-0.0261460096,
0.0440115072,
-0.0161872227,
-0.002691288,
-0.0061114007,
0.0151076429,
0.0128437793,
-0.0494498461,
-0.016648395,
-0.0386729799,
0.0117223905,
0.0461580344,
0.0220577773,
-0.0603581108,
0.0336036831,
-0.0427846909,
0.0313941576,
-0.02... |
http://arxiv.org/abs/2407.01937v1 | http://arxiv.org/pdf/2407.01937v1 | Efficient-Empathy: Towards Efficient and Effective Selection of Empathy Data | 2024-07-02T04:11:52 | 2024-07-02T04:11:52 | Linzhuang Sun, Hao Liang, Jingxuan Wei, Linkun Sun, Bihui Yu, Bin Cui, Wentao Zhang | cs.CL | cs.CL | In recent years, with the rapid advancements in large language models (LLMs),
achieving excellent empathetic response capability has become a crucial
prerequisite. Consequently, managing and understanding large-scale video
datasets has gained increasing importance. However, empathetic data are
typically trained without... | Efficient-Empathy: Towards Efficient and Effective Selection of Empathy Data: In recent years, with the rapid advancements in large language models (LLMs),
achieving excellent empathetic response capability has become a crucial
prerequisite. Consequently, managing and understanding large-scale video
datasets has gaine... | [
0.0003859997,
0.0688269809,
-0.0298611093,
0.0511481315,
0.0144834295,
0.0645776689,
-0.0127271237,
0.0068934639,
0.0102826888,
-0.0154928304,
-0.044686079,
0.0042717168,
-0.0215536375,
-0.0125465915,
0.0093397684,
-0.078232944,
0.0313323289,
0.0029406068,
0.0718985945,
-0.0251... |
http://arxiv.org/abs/2407.01929v1 | http://arxiv.org/pdf/2407.01929v1 | What We Talk About When We Talk About LMs: Implicit Paradigm Shifts and the Ship of Language Models | 2024-07-02T03:45:55 | 2024-07-02T03:45:55 | Shengqi Zhu, Jeffrey M. Rzeszotarski | cs.CL | cs.CL, cs.AI | The term Language Models (LMs), as a time-specific collection of models of
interest, is constantly reinvented, with its referents updated much like the
$\textit{Ship of Theseus}$ replaces its parts but remains the same ship in
essence. In this paper, we investigate this $\textit{Ship of Language Models}$
problem, where... | What We Talk About When We Talk About LMs: Implicit Paradigm Shifts and the Ship of Language Models: The term Language Models (LMs), as a time-specific collection of models of
interest, is constantly reinvented, with its referents updated much like the
$\textit{Ship of Theseus}$ replaces its parts but remains the same... | [
0.0472827628,
0.0654973015,
-0.0492016114,
-0.0100684613,
-0.0475820862,
-0.0266198404,
-0.0260069203,
0.0302383229,
-0.0406037383,
-0.0864335001,
-0.0072639515,
-0.0479884557,
0.0340047963,
0.0183227174,
0.0590107478,
-0.0500216931,
0.0433944874,
-0.0039661266,
-0.0019054062,
... |
http://arxiv.org/abs/2407.01920v1 | http://arxiv.org/pdf/2407.01920v1 | To Forget or Not? Towards Practical Knowledge Unlearning for Large Language Models | 2024-07-02T03:34:16 | 2024-07-02T03:34:16 | Bozhong Tian, Xiaozhuan Liang, Siyuan Cheng, Qingbin Liu, Mengru Wang, Dianbo Sui, Xi Chen, Huajun Chen, Ningyu Zhang | cs.CL | cs.CL, cs.AI, cs.CV, cs.LG, cs.MM | Large Language Models (LLMs) trained on extensive corpora inevitably retain
sensitive data, such as personal privacy information and copyrighted material.
Recent advancements in knowledge unlearning involve updating LLM parameters to
erase specific knowledge. However, current unlearning paradigms are mired in
vague for... | To Forget or Not? Towards Practical Knowledge Unlearning for Large Language Models: Large Language Models (LLMs) trained on extensive corpora inevitably retain
sensitive data, such as personal privacy information and copyrighted material.
Recent advancements in knowledge unlearning involve updating LLM parameters to
e... | [
0.0130812451,
0.1114510447,
-0.0185754038,
0.0669003502,
-0.0403508693,
0.0005724062,
0.04680245,
0.0189218372,
0.0248058755,
-0.0285734739,
-0.0617172606,
-0.0120080253,
-0.0198623557,
-0.0048369053,
-0.0118750809,
0.0128866127,
0.0319524109,
-0.0052143745,
0.0012759838,
-0.04... |
http://arxiv.org/abs/2407.01911v1 | http://arxiv.org/pdf/2407.01911v1 | Investigating the Effects of Large-Scale Pseudo-Stereo Data and Different Speech Foundation Model on Dialogue Generative Spoken Language Model | 2024-07-02T03:22:41 | 2024-07-02T03:22:41 | Yu-Kuan Fu, Cheng-Kuang Lee, Hsiu-Hsuan Wang, Hung-yi Lee | cs.CL | cs.CL, cs.HC, cs.SD, eess.AS | Recent efforts in Spoken Dialogue Modeling aim to synthesize spoken dialogue
without the need for direct transcription, thereby preserving the wealth of
non-textual information inherent in speech. However, this approach faces a
challenge when speakers talk simultaneously, requiring stereo dialogue data
with speakers re... | Investigating the Effects of Large-Scale Pseudo-Stereo Data and Different Speech Foundation Model on Dialogue Generative Spoken Language Model: Recent efforts in Spoken Dialogue Modeling aim to synthesize spoken dialogue
without the need for direct transcription, thereby preserving the wealth of
non-textual informati... | [
0.0285842288,
0.1185145155,
0.0068673999,
0.0241131689,
-0.0097646816,
-0.0090459203,
0.0272595678,
-0.0073880758,
-0.0307661258,
-0.0628768951,
-0.040223334,
-0.0694810301,
0.0332861543,
0.0658189729,
0.0349069387,
-0.0842431113,
0.0038708483,
-0.060369458,
-0.0189896785,
-0.0... |
http://arxiv.org/abs/2407.01909v1 | http://arxiv.org/pdf/2407.01909v1 | Pinyin Regularization in Error Correction for Chinese Speech Recognition with Large Language Models | 2024-07-02T03:16:47 | 2024-07-02T03:16:47 | Zhiyuan Tang, Dong Wang, Shen Huang, Shidong Shang | cs.CL | cs.CL, cs.SD, eess.AS | Recent studies have demonstrated the efficacy of large language models (LLMs)
in error correction for automatic speech recognition (ASR). However, much of
the research focuses on the English language. This paper redirects the
attention to Chinese. Firstly, we construct a specialized benchmark dataset
aimed at error cor... | Pinyin Regularization in Error Correction for Chinese Speech Recognition with Large Language Models: Recent studies have demonstrated the efficacy of large language models (LLMs)
in error correction for automatic speech recognition (ASR). However, much of
the research focuses on the English language. This paper redire... | [
0.0137403607,
0.0659070015,
0.0297465213,
0.0290342942,
0.0149502261,
0.0082852142,
0.0435149483,
0.0251533277,
0.0073620742,
0.0069242939,
-0.0815744773,
-0.0755984038,
0.0416900031,
-0.0417638943,
0.0305635948,
-0.0117005687,
0.0329296961,
-0.0432014689,
0.0016000272,
-0.0270... |
http://arxiv.org/abs/2407.01906v1 | http://arxiv.org/pdf/2407.01906v1 | Let the Expert Stick to His Last: Expert-Specialized Fine-Tuning for Sparse Architectural Large Language Models | 2024-07-02T03:11:13 | 2024-07-02T03:11:13 | Zihan Wang, Deli Chen, Damai Dai, Runxin Xu, Zhuoshu Li, Y. Wu | cs.CL | cs.CL, cs.AI, cs.LG | Parameter-efficient fine-tuning (PEFT) is crucial for customizing Large
Language Models (LLMs) with constrained resources. Although there have been
various PEFT methods for dense-architecture LLMs, PEFT for sparse-architecture
LLMs is still underexplored. In this work, we study the PEFT method for LLMs
with the Mixture... | Let the Expert Stick to His Last: Expert-Specialized Fine-Tuning for Sparse Architectural Large Language Models: Parameter-efficient fine-tuning (PEFT) is crucial for customizing Large
Language Models (LLMs) with constrained resources. Although there have been
various PEFT methods for dense-architecture LLMs, PEFT for... | [
0.0632905215,
0.0403175093,
-0.020782914,
0.0333820134,
-0.0110380305,
-0.0064116283,
0.0223335698,
-0.0290605836,
-0.0371697433,
-0.0437917113,
-0.0694279969,
-0.0134273656,
0.0003318896,
-0.008011709,
0.05341167,
-0.0092549706,
0.0013880415,
-0.0594246089,
-0.0029778297,
-0.0... |
http://arxiv.org/abs/2407.01902v1 | http://arxiv.org/pdf/2407.01902v1 | SoP: Unlock the Power of Social Facilitation for Automatic Jailbreak Attack | 2024-07-02T02:58:29 | 2024-07-02T02:58:29 | Yan Yang, Zeguan Xiao, Xin Lu, Hongru Wang, Hailiang Huang, Guanhua Chen, Yun Chen | cs.CR | cs.CR, cs.AI, cs.CL | The widespread applications of large language models (LLMs) have brought
about concerns regarding their potential misuse. Although aligned with human
preference data before release, LLMs remain vulnerable to various malicious
attacks. In this paper, we adopt a red-teaming strategy to enhance LLM safety
and introduce So... | SoP: Unlock the Power of Social Facilitation for Automatic Jailbreak Attack: The widespread applications of large language models (LLMs) have brought
about concerns regarding their potential misuse. Although aligned with human
preference data before release, LLMs remain vulnerable to various malicious
attacks. In this... | [
0.0691232011,
0.0390567854,
-0.0111071942,
0.008895454,
-0.0253803115,
-0.0335569419,
0.0326035917,
-0.016866453,
0.0514845736,
-0.0084161628,
-0.0147852302,
-0.0171891078,
0.0251117703,
0.006804165,
0.0346463472,
0.0128526241,
0.0223340373,
-0.0254952852,
-0.0043451376,
0.0107... |
http://arxiv.org/abs/2407.01899v1 | http://arxiv.org/pdf/2407.01899v1 | Scope-enhanced Compositional Semantic Parsing for DRT | 2024-07-02T02:50:15 | 2024-07-02T02:50:15 | Xiulin Yang, Jonas Groschwitz, Alexander Koller, Johan Bos | cs.CL | cs.CL | Discourse Representation Theory (DRT) distinguishes itself from other
semantic representation frameworks by its ability to model complex semantic and
discourse phenomena through structural nesting and variable binding. While
seq2seq models hold the state of the art on DRT parsing, their accuracy
degrades with the compl... | Scope-enhanced Compositional Semantic Parsing for DRT: Discourse Representation Theory (DRT) distinguishes itself from other
semantic representation frameworks by its ability to model complex semantic and
discourse phenomena through structural nesting and variable binding. While
seq2seq models hold the state of the art... | [
0.0030659393,
0.0695008561,
-0.033425089,
0.0161633659,
-0.0188846737,
0.0116902944,
-0.023353558,
-0.0347311012,
-0.0242804177,
-0.0695567876,
-0.0256995652,
0.0222253166,
-0.0314985141,
0.0367292799,
-0.0031011994,
-0.0465509333,
0.077458024,
-0.0282822978,
0.0355019234,
0.01... |
http://arxiv.org/abs/2407.01897v1 | http://arxiv.org/pdf/2407.01897v1 | Proposal Report for the 2nd SciCAP Competition 2024 | 2024-07-02T02:42:29 | 2024-07-02T02:42:29 | Pengpeng Li, Tingmin Li, Jingyuan Wang, Boyuan Wang, Yang Yang | cs.CL | cs.CL | In this paper, we propose a method for document summarization using auxiliary
information. This approach effectively summarizes descriptions related to
specific images, tables, and appendices within lengthy texts. Our experiments
demonstrate that leveraging high-quality OCR data and initially extracted
information from... | Proposal Report for the 2nd SciCAP Competition 2024: In this paper, we propose a method for document summarization using auxiliary
information. This approach effectively summarizes descriptions related to
specific images, tables, and appendices within lengthy texts. Our experiments
demonstrate that leveraging high-qual... | [
0.0344179757,
0.0368236154,
0.0440982357,
-0.0148424814,
-0.0053793294,
0.0173082668,
-0.0279916674,
0.0380046293,
-0.0162920486,
-0.0514986366,
0.0168521758,
0.0112207215,
0.0472348221,
0.0184814706,
0.0100370543,
-0.1041243896,
-0.0013667557,
-0.0497315265,
-0.0362303816,
0.0... |
http://arxiv.org/abs/2407.01896v1 | http://arxiv.org/pdf/2407.01896v1 | LogEval: A Comprehensive Benchmark Suite for Large Language Models In Log Analysis | 2024-07-02T02:39:33 | 2024-07-02T02:39:33 | Tianyu Cui, Shiyu Ma, Ziang Chen, Tong Xiao, Shimin Tao, Yilun Liu, Shenglin Zhang, Duoming Lin, Changchang Liu, Yuzhe Cai, Weibin Meng, Yongqian Sun, Dan Pei | cs.CL | cs.CL, cs.IR | Log analysis is crucial for ensuring the orderly and stable operation of
information systems, particularly in the field of Artificial Intelligence for
IT Operations (AIOps). Large Language Models (LLMs) have demonstrated
significant potential in natural language processing tasks. In the AIOps
domain, they excel in task... | LogEval: A Comprehensive Benchmark Suite for Large Language Models In Log Analysis: Log analysis is crucial for ensuring the orderly and stable operation of
information systems, particularly in the field of Artificial Intelligence for
IT Operations (AIOps). Large Language Models (LLMs) have demonstrated
significant po... | [
0.0537884012,
0.0216487255,
0.0109076044,
0.017958479,
-0.0399849191,
-0.0014257265,
0.0281325821,
-0.0278127361,
0.0059421333,
-0.068236798,
0.0269456562,
-0.0704207271,
-0.0456926674,
0.0696982294,
-0.0074957064,
-0.0147964889,
0.0589918867,
-0.0064309156,
-0.0225287098,
0.02... |
http://arxiv.org/abs/2407.01892v1 | http://arxiv.org/pdf/2407.01892v1 | GRASP: A Grid-Based Benchmark for Evaluating Commonsense Spatial Reasoning | 2024-07-02T02:27:46 | 2024-07-02T02:27:46 | Zhisheng Tang, Mayank Kejriwal | cs.AI | cs.AI, cs.CL | Spatial reasoning, an important faculty of human cognition with many
practical applications, is one of the core commonsense skills that is not
purely language-based and, for satisfying (as opposed to optimal) solutions,
requires some minimum degree of planning. Existing benchmarks of Commonsense
Spatial Reasoning (CSR)... | GRASP: A Grid-Based Benchmark for Evaluating Commonsense Spatial Reasoning: Spatial reasoning, an important faculty of human cognition with many
practical applications, is one of the core commonsense skills that is not
purely language-based and, for satisfying (as opposed to optimal) solutions,
requires some minimum d... | [
0.0559374094,
0.0298450086,
-0.003924605,
0.030710509,
-0.0534627698,
-0.0078474265,
-0.0135781514,
-0.0089689586,
0.0645890236,
-0.0007486281,
-0.0029217941,
0.0114600221,
0.0280184224,
-0.0292919986,
0.0121481083,
-0.051919315,
0.0221699458,
-0.0393278301,
-0.0379963815,
-0.0... |
http://arxiv.org/abs/2407.01887v1 | http://arxiv.org/pdf/2407.01887v1 | Beyond Numeric Awards: In-Context Dueling Bandits with LLM Agents | 2024-07-02T02:18:14 | 2024-07-02T02:18:14 | Fanzeng Xia, Hao Liu, Yisong Yue, Tongxin Li | cs.LG | cs.LG, cs.AI, cs.CL | In-context decision-making is an important capability of artificial general
intelligence, which Large Language Models (LLMs) have effectively demonstrated
in various scenarios. However, LLMs often face challenges when dealing with
numerical contexts, and limited attention has been paid to evaluating their
performance t... | Beyond Numeric Awards: In-Context Dueling Bandits with LLM Agents: In-context decision-making is an important capability of artificial general
intelligence, which Large Language Models (LLMs) have effectively demonstrated
in various scenarios. However, LLMs often face challenges when dealing with
numerical contexts, an... | [
0.036559213,
0.0600022934,
0.0018728066,
0.0418717004,
-0.0670031086,
0.0098423036,
-0.0278959516,
0.0127221057,
0.0357890017,
-0.0258210842,
-0.0484823063,
-0.0449298322,
-0.0219582934,
-0.0075206035,
0.0210883562,
0.0033905967,
0.0357822292,
-0.0706925169,
0.0532721616,
-0.04... |
http://arxiv.org/abs/2407.01885v1 | http://arxiv.org/pdf/2407.01885v1 | Survey on Knowledge Distillation for Large Language Models: Methods, Evaluation, and Application | 2024-07-02T02:14:42 | 2024-07-02T02:14:42 | Chuanpeng Yang, Wang Lu, Yao Zhu, Yidong Wang, Qian Chen, Chenlong Gao, Bingjie Yan, Yiqiang Chen | cs.CL | cs.CL, cs.AI | Large Language Models (LLMs) have showcased exceptional capabilities in
various domains, attracting significant interest from both academia and
industry. Despite their impressive performance, the substantial size and
computational demands of LLMs pose considerable challenges for practical
deployment, particularly in en... | Survey on Knowledge Distillation for Large Language Models: Methods, Evaluation, and Application: Large Language Models (LLMs) have showcased exceptional capabilities in
various domains, attracting significant interest from both academia and
industry. Despite their impressive performance, the substantial size and
comp... | [
0.0059897946,
0.0860266536,
-0.0261540916,
0.0655396953,
-0.0641787499,
0.019915171,
-0.0127922222,
0.0374399014,
-0.0180339646,
0.0032139414,
0.0002352078,
-0.035886135,
0.0227394197,
0.0069933408,
0.0140310712,
-0.0129690636,
0.0210626535,
-0.0681547672,
-0.0016568387,
-0.023... |
http://arxiv.org/abs/2407.01878v1 | http://arxiv.org/pdf/2407.01878v1 | Compare without Despair: Reliable Preference Evaluation with Generation Separability | 2024-07-02T01:37:56 | 2024-07-02T01:37:56 | Sayan Ghosh, Tejas Srinivasan, Swabha Swayamdipta | cs.CL | cs.CL | Human evaluation of generated language through pairwise preference judgments
is pervasive. However, under common scenarios, such as when generations from a
model pair are very similar, or when stochastic decoding results in large
variations in generations, it results in inconsistent preference ratings. We
address these... | Compare without Despair: Reliable Preference Evaluation with Generation Separability: Human evaluation of generated language through pairwise preference judgments
is pervasive. However, under common scenarios, such as when generations from a
model pair are very similar, or when stochastic decoding results in large
var... | [
-0.0062347255,
0.0887169093,
-0.0093275784,
0.0305363778,
-0.0278459918,
0.0090485234,
0.0102385478,
0.0040825708,
-0.0401499346,
-0.030423522,
-0.0736105219,
-0.031591136,
0.0046233693,
0.0298977531,
0.0009454543,
-0.0417149141,
0.0318351723,
-0.0243537836,
0.0170984343,
-0.01... |
http://arxiv.org/abs/2407.01873v1 | http://arxiv.org/pdf/2407.01873v1 | Automated Text Scoring in the Age of Generative AI for the GPU-poor | 2024-07-02T01:17:01 | 2024-07-02T01:17:01 | Christopher Michael Ormerod, Alexander Kwako | cs.LG | cs.LG, cs.AI, cs.CL | Current research on generative language models (GLMs) for automated text
scoring (ATS) has focused almost exclusively on querying proprietary models via
Application Programming Interfaces (APIs). Yet such practices raise issues
around transparency and security, and these methods offer little in the way of
efficiency or... | Automated Text Scoring in the Age of Generative AI for the GPU-poor: Current research on generative language models (GLMs) for automated text
scoring (ATS) has focused almost exclusively on querying proprietary models via
Application Programming Interfaces (APIs). Yet such practices raise issues
around transparency and... | [
0.0273657888,
0.0795835257,
-0.0073528597,
0.0206906591,
-0.0483401343,
-0.0016400812,
0.0325370803,
0.0068389717,
0.028090965,
-0.0459867977,
0.0062424629,
-0.034218505,
0.0511112586,
0.0554534532,
0.0179953035,
-0.0203939714,
-0.0033587727,
-0.0249401964,
0.0276788604,
-0.024... |
http://arxiv.org/abs/2407.01863v1 | http://arxiv.org/pdf/2407.01863v1 | VSP: Assessing the dual challenges of perception and reasoning in spatial planning tasks for VLMs | 2024-07-02T00:24:01 | 2024-07-02T00:24:01 | Qiucheng Wu, Handong Zhao, Michael Saxon, Trung Bui, William Yang Wang, Yang Zhang, Shiyu Chang | cs.CL | cs.CL | Vision language models (VLMs) are an exciting emerging class of language
models (LMs) that have merged classic LM capabilities with those of image
processing systems. However, the ways that these capabilities combine are not
always intuitive and warrant direct investigation. One understudied capability
in VLMs is visua... | VSP: Assessing the dual challenges of perception and reasoning in spatial planning tasks for VLMs: Vision language models (VLMs) are an exciting emerging class of language
models (LMs) that have merged classic LM capabilities with those of image
processing systems. However, the ways that these capabilities combine are... | [
0.0507315435,
0.0419118106,
-0.0108187459,
0.0401357971,
-0.0749728978,
0.0102303568,
0.0193182137,
-0.0241470747,
0.0279957317,
-0.0407124236,
-0.0074548521,
0.0016785293,
0.0409332551,
0.0401141979,
-0.0164330788,
-0.0641214773,
0.0072461627,
-0.0680598617,
-0.008654654,
-0.0... |
http://arxiv.org/abs/2407.01853v1 | http://arxiv.org/pdf/2407.01853v1 | Improving Multilingual Instruction Finetuning via Linguistically Natural and Diverse Datasets | 2024-07-01T23:47:09 | 2024-07-01T23:47:09 | Sathish Reddy Indurthi, Wenxuan Zhou, Shamil Chollampatt, Ravi Agrawal, Kaiqiang Song, Lingxiao Zhao, Chenguang Zhu | cs.CL | cs.CL, cs.AI, cs.LG | Advancements in Large Language Models (LLMs) have significantly enhanced
instruction-following capabilities. However, most Instruction Fine-Tuning (IFT)
datasets are predominantly in English, limiting model performance in other
languages. Traditional methods for creating multilingual IFT datasets such as
translating ex... | Improving Multilingual Instruction Finetuning via Linguistically Natural and Diverse Datasets: Advancements in Large Language Models (LLMs) have significantly enhanced
instruction-following capabilities. However, most Instruction Fine-Tuning (IFT)
datasets are predominantly in English, limiting model performance in ot... | [
0.033930514,
0.0163812321,
-0.0095963068,
0.0238033906,
-0.0179088172,
0.0292194262,
0.0278173033,
-0.0617413037,
-0.0132535445,
-0.0320652835,
-0.0510458611,
-0.0132690296,
0.0053487262,
-0.0399073549,
0.0359891541,
-0.0563021898,
0.0447623916,
-0.0373049267,
-0.0442342311,
-0... |
http://arxiv.org/abs/2407.01850v1 | http://arxiv.org/pdf/2407.01850v1 | Purple-teaming LLMs with Adversarial Defender Training | 2024-07-01T23:25:30 | 2024-07-01T23:25:30 | Jingyan Zhou, Kun Li, Junan Li, Jiawen Kang, Minda Hu, Xixin Wu, Helen Meng | cs.CL | cs.CL | Existing efforts in safeguarding LLMs are limited in actively exposing the
vulnerabilities of the target LLM and readily adapting to newly emerging safety
risks. To address this, we present Purple-teaming LLMs with Adversarial
Defender training (PAD), a pipeline designed to safeguard LLMs by novelly
incorporating the r... | Purple-teaming LLMs with Adversarial Defender Training: Existing efforts in safeguarding LLMs are limited in actively exposing the
vulnerabilities of the target LLM and readily adapting to newly emerging safety
risks. To address this, we present Purple-teaming LLMs with Adversarial
Defender training (PAD), a pipeline d... | [
-0.016712103,
0.0664671808,
-0.001861298,
0.0279000401,
-0.0323689505,
-0.0335445069,
0.0162320118,
-0.0068892613,
-0.0233422574,
-0.0167471915,
-0.0156459734,
0.0143274339,
-0.0204088092,
-0.0025695572,
0.0812278762,
-0.0117646679,
0.0709948838,
-0.0357602388,
-0.0188968834,
0... |
http://arxiv.org/abs/2407.01834v1 | http://arxiv.org/pdf/2407.01834v1 | A Study of Nationality Bias in Names and Perplexity using Off-the-Shelf Affect-related Tweet Classifiers | 2024-07-01T22:17:17 | 2024-07-01T22:17:17 | Valentin Barriere, Sebastian Cifuentes | cs.CL | cs.CL | In this paper, we apply a method to quantify biases associated with named
entities from various countries. We create counterfactual examples with small
perturbations on target-domain data instead of relying on templates or specific
datasets for bias detection. On widely used classifiers for subjectivity
analysis, inclu... | A Study of Nationality Bias in Names and Perplexity using Off-the-Shelf Affect-related Tweet Classifiers: In this paper, we apply a method to quantify biases associated with named
entities from various countries. We create counterfactual examples with small
perturbations on target-domain data instead of relying on tem... | [
0.0443994552,
0.1055501699,
-0.0358166099,
0.014002393,
-0.0109988451,
0.0657564402,
0.0411430746,
0.0014062582,
0.0421725996,
-0.025489457,
0.0127323344,
-0.0147935459,
-0.0138417603,
-0.009905993,
-0.0019473807,
-0.0604913384,
0.0224418007,
0.0548605062,
0.061066132,
-0.05702... |
http://arxiv.org/abs/2407.01824v1 | http://arxiv.org/pdf/2407.01824v1 | Empathic Grounding: Explorations using Multimodal Interaction and Large Language Models with Conversational Agents | 2024-07-01T21:46:30 | 2024-07-01T21:46:30 | Mehdi Arjmand, Farnaz Nouraei, Ian Steenstra, Timothy Bickmore | cs.HC | cs.HC, cs.CL, cs.RO | We introduce the concept of "empathic grounding" in conversational agents as
an extension of Clark's conceptualization of grounding in conversation in which
the grounding criterion includes listener empathy for the speaker's affective
state. Empathic grounding is generally required whenever the speaker's emotions
are f... | Empathic Grounding: Explorations using Multimodal Interaction and Large Language Models with Conversational Agents: We introduce the concept of "empathic grounding" in conversational agents as
an extension of Clark's conceptualization of grounding in conversation in which
the grounding criterion includes listener empa... | [
0.043619372,
0.0007324757,
-0.0209490191,
-0.0031037519,
0.0000152796,
-0.0102066537,
0.0463763624,
-0.0073916367,
0.01349092,
-0.0204455592,
-0.0397858024,
0.0221933108,
-0.0010412837,
0.0041144565,
0.032321617,
-0.0709361956,
0.0041658212,
0.0327020511,
0.0612528995,
0.020757... |
http://arxiv.org/abs/2407.01817v1 | http://arxiv.org/pdf/2407.01817v1 | Race and Privacy in Broadcast Police Communications | 2024-07-01T21:34:51 | 2024-07-01T21:34:51 | Pranav Narayanan Venkit, Christopher Graziul, Miranda Ardith Goodman, Samantha Nicole Kenny, Shomir Wilson | cs.CL | cs.CL, cs.CY, cs.HC | Radios are essential for the operations of modern police departments, and
they function as both a collaborative communication technology and a
sociotechnical system. However, little prior research has examined their usage
or their connections to individual privacy and the role of race in policing,
two growing topics of... | Race and Privacy in Broadcast Police Communications: Radios are essential for the operations of modern police departments, and
they function as both a collaborative communication technology and a
sociotechnical system. However, little prior research has examined their usage
or their connections to individual privacy an... | [
0.0325931422,
0.0732274652,
-0.0247501843,
-0.0185038168,
-0.0557259955,
0.0332426727,
0.0763183832,
-0.0295511205,
-0.0440677106,
-0.0453169905,
-0.0044904803,
-0.0256020483,
0.0271223616,
-0.0335011072,
-0.0402890891,
-0.0139927519,
-0.003676825,
0.0034284438,
0.0498910472,
0... |
http://arxiv.org/abs/2407.01796v1 | http://arxiv.org/pdf/2407.01796v1 | Ground Every Sentence: Improving Retrieval-Augmented LLMs with Interleaved Reference-Claim Generation | 2024-07-01T20:47:47 | 2024-07-01T20:47:47 | Sirui Xia, Xintao Wang, Jiaqing Liang, Yifei Zhang, Weikang Zhou, Jiaji Deng, Fei Yu, Yanghua Xiao | cs.CL | cs.CL | Retrieval-Augmented Generation (RAG) has been widely adopted to enhance Large
Language Models (LLMs) in knowledge-intensive tasks. Recently, Attributed Text
Generation (ATG) has attracted growing attention, which provides citations to
support the model's responses in RAG, so as to enhance the credibility of
LLM-generat... | Ground Every Sentence: Improving Retrieval-Augmented LLMs with Interleaved Reference-Claim Generation: Retrieval-Augmented Generation (RAG) has been widely adopted to enhance Large
Language Models (LLMs) in knowledge-intensive tasks. Recently, Attributed Text
Generation (ATG) has attracted growing attention, which pro... | [
0.0617013425,
0.0186267328,
0.0150335301,
0.0230915919,
-0.0333398357,
-0.0201109983,
-0.0027793655,
-0.0317353643,
0.0072811861,
-0.088499397,
0.0025862707,
0.0149748176,
-0.0303230025,
-0.061738532,
0.058577314,
0.0189862028,
0.0374314152,
0.0017143787,
0.0253855493,
-0.00304... |
http://arxiv.org/abs/2407.01784v1 | http://arxiv.org/pdf/2407.01784v1 | Analyzing Persuasive Strategies in Meme Texts: A Fusion of Language Models with Paraphrase Enrichment | 2024-07-01T20:25:20 | 2024-07-01T20:25:20 | Kota Shamanth Ramanath Nayak, Leila Kosseim | cs.CL | cs.CL, cs.AI, cs.LG | This paper describes our approach to hierarchical multi-label detection of
persuasion techniques in meme texts. Our model, developed as a part of the
recent SemEval task, is based on fine-tuning individual language models (BERT,
XLM-RoBERTa, and mBERT) and leveraging a mean-based ensemble model in addition
to dataset a... | Analyzing Persuasive Strategies in Meme Texts: A Fusion of Language Models with Paraphrase Enrichment: This paper describes our approach to hierarchical multi-label detection of
persuasion techniques in meme texts. Our model, developed as a part of the
recent SemEval task, is based on fine-tuning individual language m... | [
0.0653997809,
0.0338543989,
-0.019100748,
0.0311046038,
-0.0319484882,
0.0208664984,
0.001108685,
0.012536101,
0.0409652069,
-0.0733941421,
-0.0317793079,
-0.0075955787,
-0.007198629,
0.0030956168,
0.0355277583,
-0.0598327145,
-0.0044777277,
0.0060986509,
0.0639885515,
-0.02873... |
http://arxiv.org/abs/2407.01725v1 | http://arxiv.org/pdf/2407.01725v1 | DiscoveryBench: Towards Data-Driven Discovery with Large Language Models | 2024-07-01T18:58:22 | 2024-07-01T18:58:22 | Bodhisattwa Prasad Majumder, Harshit Surana, Dhruv Agarwal, Bhavana Dalvi Mishra, Abhijeetsingh Meena, Aryan Prakhar, Tirth Vora, Tushar Khot, Ashish Sabharwal, Peter Clark | cs.CL | cs.CL, cs.AI, cs.LG | Can the rapid advances in code generation, function calling, and data
analysis using large language models (LLMs) help automate the search and
verification of hypotheses purely from a set of provided datasets? To evaluate
this question, we present DiscoveryBench, the first comprehensive benchmark
that formalizes the mu... | DiscoveryBench: Towards Data-Driven Discovery with Large Language Models: Can the rapid advances in code generation, function calling, and data
analysis using large language models (LLMs) help automate the search and
verification of hypotheses purely from a set of provided datasets? To evaluate
this question, we presen... | [
-0.0055923657,
0.0726017952,
-0.0218305271,
0.0401221775,
-0.0445148088,
-0.0121522006,
0.0010594911,
-0.0121931201,
0.0376392379,
-0.016695274,
0.0323835462,
0.0160483345,
-0.0187014937,
0.0537822656,
-0.0132696023,
0.0200655572,
0.0388038084,
-0.0344077908,
0.0043742252,
0.00... |
http://arxiv.org/abs/2407.01697v1 | http://arxiv.org/pdf/2407.01697v1 | NLPGuard: A Framework for Mitigating the Use of Protected Attributes by NLP Classifiers | 2024-07-01T18:08:17 | 2024-07-01T18:08:17 | Salvatore Greco, Ke Zhou, Licia Capra, Tania Cerquitelli, Daniele Quercia | cs.CL | cs.CL, cs.AI, cs.HC | AI regulations are expected to prohibit machine learning models from using
sensitive attributes during training. However, the latest Natural Language
Processing (NLP) classifiers, which rely on deep learning, operate as black-box
systems, complicating the detection and remediation of such misuse. Traditional
bias mitig... | NLPGuard: A Framework for Mitigating the Use of Protected Attributes by NLP Classifiers: AI regulations are expected to prohibit machine learning models from using
sensitive attributes during training. However, the latest Natural Language
Processing (NLP) classifiers, which rely on deep learning, operate as black-box
... | [
0.0196209438,
0.1353103518,
-0.0150950737,
0.0132507728,
-0.0178322569,
0.0235601831,
0.0382625014,
-0.0159020666,
0.0104663828,
-0.0753031,
-0.0053235823,
-0.0206115916,
-0.0034613274,
0.0309144203,
-0.0013386769,
-0.0167269353,
0.0300701093,
0.0417619161,
0.0470651612,
-0.021... |
http://arxiv.org/abs/2407.01687v1 | http://arxiv.org/pdf/2407.01687v1 | Deciphering the Factors Influencing the Efficacy of Chain-of-Thought: Probability, Memorization, and Noisy Reasoning | 2024-07-01T18:01:07 | 2024-07-01T18:01:07 | Akshara Prabhakar, Thomas L. Griffiths, R. Thomas McCoy | cs.CL | cs.CL, cs.AI | Chain-of-Thought (CoT) prompting has been shown to enhance the multi-step
reasoning capabilities of Large Language Models (LLMs). However, debates
persist about whether LLMs exhibit abstract generalization or rely on shallow
heuristics when given CoT prompts. To understand the factors influencing CoT
reasoning we provi... | Deciphering the Factors Influencing the Efficacy of Chain-of-Thought: Probability, Memorization, and Noisy Reasoning: Chain-of-Thought (CoT) prompting has been shown to enhance the multi-step
reasoning capabilities of Large Language Models (LLMs). However, debates
persist about whether LLMs exhibit abstract generaliza... | [
0.0253911167,
0.0642234981,
-0.0100400429,
0.0153231146,
-0.0800844729,
0.0447710007,
0.0380543843,
-0.0137039237,
-0.0209676251,
0.0060209958,
0.0223184098,
-0.0268468335,
0.0332106799,
-0.0130218789,
0.0336554348,
-0.0283980351,
0.0338959247,
-0.1059884578,
-0.0056705591,
-0.... |
http://arxiv.org/abs/2407.01527v1 | http://arxiv.org/pdf/2407.01527v1 | KV Cache Compression, But What Must We Give in Return? A Comprehensive Benchmark of Long Context Capable Approaches | 2024-07-01T17:59:47 | 2024-07-01T17:59:47 | Jiayi Yuan, Hongyi Liu, Shaochen, Zhong, Yu-Neng Chuang, Songchen Li, Guanchu Wang, Duy Le, Hongye Jin, Vipin Chaudhary, Zhaozhuo Xu, Zirui Liu, Xia Hu | cs.CL | cs.CL | Long context capability is a crucial competency for large language models
(LLMs) as it mitigates the human struggle to digest long-form texts. This
capability enables complex task-solving scenarios such as book summarization,
code assistance, and many more tasks that are traditionally manpower-intensive.
However, trans... | KV Cache Compression, But What Must We Give in Return? A Comprehensive Benchmark of Long Context Capable Approaches: Long context capability is a crucial competency for large language models
(LLMs) as it mitigates the human struggle to digest long-form texts. This
capability enables complex task-solving scenarios such... | [
0.0150038386,
0.0224205684,
-0.0134197501,
0.0276555214,
-0.0479249619,
0.0114462487,
0.0134575535,
0.0230105426,
-0.0145785511,
-0.0284229461,
0.0037904815,
-0.0445857979,
0.0169695951,
0.0149822803,
0.0171172097,
-0.0202520285,
0.0474628247,
-0.0412464552,
-0.0221970882,
-0.0... |
http://arxiv.org/abs/2407.01525v2 | http://arxiv.org/pdf/2407.01525v2 | Empowering 3D Visual Grounding with Reasoning Capabilities | 2024-07-01T17:59:35 | 2024-07-02T07:37:56 | Chenming Zhu, Tai Wang, Wenwei Zhang, Kai Chen, Xihui Liu | cs.CV | cs.CV, cs.AI, cs.CL | Although great progress has been made in 3D visual grounding, current models
still rely on explicit textual descriptions for grounding and lack the ability
to reason human intentions from implicit instructions. We propose a new task
called 3D reasoning grounding and introduce a new benchmark ScanReason which
provides o... | Empowering 3D Visual Grounding with Reasoning Capabilities: Although great progress has been made in 3D visual grounding, current models
still rely on explicit textual descriptions for grounding and lack the ability
to reason human intentions from implicit instructions. We propose a new task
called 3D reasoning groundi... | [
0.0356089547,
0.040079914,
0.0366839021,
0.0495455153,
-0.0176378358,
0.0124703012,
0.0274359211,
-0.020247858,
-0.0178444404,
-0.0651834607,
-0.0025057246,
0.0231088176,
0.020177044,
0.0266983565,
0.0344130434,
-0.0501927286,
0.0070290198,
-0.0469935946,
-0.0163271409,
-0.0002... |
http://arxiv.org/abs/2407.01523v1 | http://arxiv.org/pdf/2407.01523v1 | MMLongBench-Doc: Benchmarking Long-context Document Understanding with Visualizations | 2024-07-01T17:59:26 | 2024-07-01T17:59:26 | Yubo Ma, Yuhang Zang, Liangyu Chen, Meiqi Chen, Yizhu Jiao, Xinze Li, Xinyuan Lu, Ziyu Liu, Yan Ma, Xiaoyi Dong, Pan Zhang, Liangming Pan, Yu-Gang Jiang, Jiaqi Wang, Yixin Cao, Aixin Sun | cs.CV | cs.CV, cs.CL | Understanding documents with rich layouts and multi-modal components is a
long-standing and practical task. Recent Large Vision-Language Models (LVLMs)
have made remarkable strides in various tasks, particularly in single-page
document understanding (DU). However, their abilities on long-context DU remain
an open probl... | MMLongBench-Doc: Benchmarking Long-context Document Understanding with Visualizations: Understanding documents with rich layouts and multi-modal components is a
long-standing and practical task. Recent Large Vision-Language Models (LVLMs)
have made remarkable strides in various tasks, particularly in single-page
docum... | [
-0.0187474582,
0.0135533158,
0.016952103,
0.0516976677,
-0.0390733071,
0.0323070027,
0.0123499231,
0.0194106288,
0.009160134,
-0.0192536283,
-0.0240349881,
-0.0242383722,
0.0185595136,
0.0232973099,
0.0113444533,
-0.0495962426,
0.0212984271,
-0.071828872,
-0.0138182007,
0.03234... |
http://arxiv.org/abs/2407.01509v1 | http://arxiv.org/pdf/2407.01509v1 | MIA-Bench: Towards Better Instruction Following Evaluation of Multimodal LLMs | 2024-07-01T17:53:35 | 2024-07-01T17:53:35 | Yusu Qian, Hanrong Ye, Jean-Philippe Fauconnier, Peter Grasch, Yinfei Yang, Zhe Gan | cs.CV | cs.CV, cs.CL | We introduce MIA-Bench, a new benchmark designed to evaluate multimodal large
language models (MLLMs) on their ability to strictly adhere to complex
instructions. Our benchmark comprises a diverse set of 400 image-prompt pairs,
each crafted to challenge the models' compliance with layered instructions in
generating acc... | MIA-Bench: Towards Better Instruction Following Evaluation of Multimodal LLMs: We introduce MIA-Bench, a new benchmark designed to evaluate multimodal large
language models (MLLMs) on their ability to strictly adhere to complex
instructions. Our benchmark comprises a diverse set of 400 image-prompt pairs,
each crafted... | [
0.018101627,
0.0636366904,
-0.007986974,
0.0563669428,
-0.0384668447,
0.0108244596,
0.0318387598,
-0.0042928299,
0.0086560277,
-0.0668320507,
-0.071742408,
-0.0510750152,
0.0065844567,
-0.0059390482,
0.04308163,
-0.0452397428,
0.0216256697,
-0.0759559348,
-0.0526828021,
-0.0140... |
http://arxiv.org/abs/2407.01505v1 | http://arxiv.org/pdf/2407.01505v1 | Self-Cognition in Large Language Models: An Exploratory Study | 2024-07-01T17:52:05 | 2024-07-01T17:52:05 | Dongping Chen, Jiawen Shi, Yao Wan, Pan Zhou, Neil Zhenqiang Gong, Lichao Sun | cs.CL | cs.CL, cs.AI | While Large Language Models (LLMs) have achieved remarkable success across
various applications, they also raise concerns regarding self-cognition. In
this paper, we perform a pioneering study to explore self-cognition in LLMs.
Specifically, we first construct a pool of self-cognition instruction prompts
to evaluate wh... | Self-Cognition in Large Language Models: An Exploratory Study: While Large Language Models (LLMs) have achieved remarkable success across
various applications, they also raise concerns regarding self-cognition. In
this paper, we perform a pioneering study to explore self-cognition in LLMs.
Specifically, we first constr... | [
0.0512771755,
0.0077245249,
-0.0471525192,
0.015218785,
-0.0446907021,
-0.0022465619,
0.026449766,
-0.0000194612,
0.0703866258,
-0.034615092,
-0.0695911869,
0.0082665849,
-0.0193744432,
0.0123597197,
0.0367696956,
-0.0276856013,
0.0410159789,
-0.0490167066,
0.0316607393,
-0.017... |
http://arxiv.org/abs/2407.01492v1 | http://arxiv.org/pdf/2407.01492v1 | RegMix: Data Mixture as Regression for Language Model Pre-training | 2024-07-01T17:31:03 | 2024-07-01T17:31:03 | Qian Liu, Xiaosen Zheng, Niklas Muennighoff, Guangtao Zeng, Longxu Dou, Tianyu Pang, Jing Jiang, Min Lin | cs.CL | cs.CL, cs.AI | The data mixture for large language model pre-training significantly impacts
performance, yet how to determine an effective mixture remains unclear. We
propose RegMix to automatically identify a high-performing data mixture by
formulating it as a regression task. RegMix involves training a set of small
models with dive... | RegMix: Data Mixture as Regression for Language Model Pre-training: The data mixture for large language model pre-training significantly impacts
performance, yet how to determine an effective mixture remains unclear. We
propose RegMix to automatically identify a high-performing data mixture by
formulating it as a regre... | [
0.0226724185,
0.0829120576,
-0.0250786673,
0.0333399363,
0.0098517872,
0.0269735027,
0.0268867556,
0.044978749,
-0.0108481525,
-0.0780734867,
-0.0174798165,
-0.0343782753,
0.0239845328,
0.0331923403,
0.0378653184,
-0.0773756951,
0.0393526293,
-0.0478743128,
-0.0088364109,
-0.01... |
http://arxiv.org/abs/2407.01491v1 | http://arxiv.org/pdf/2407.01491v1 | Expressive and Generalizable Low-rank Adaptation for Large Models via Slow Cascaded Learning | 2024-07-01T17:28:59 | 2024-07-01T17:28:59 | Siwei Li, Yifan Yang, Yifei Shen, Fangyun Wei, Zongqing Lu, Lili Qiu, Yuqing Yang | cs.CL | cs.CL, cs.CV | Efficient fine-tuning plays a fundamental role in modern large models, with
low-rank adaptation emerging as a particularly promising approach. However, the
existing variants of LoRA are hampered by limited expressiveness, a tendency to
overfit, and sensitivity to hyperparameter settings. This paper presents LoRA
Slow C... | Expressive and Generalizable Low-rank Adaptation for Large Models via Slow Cascaded Learning: Efficient fine-tuning plays a fundamental role in modern large models, with
low-rank adaptation emerging as a particularly promising approach. However, the
existing variants of LoRA are hampered by limited expressiveness, a t... | [
0.0002231044,
0.1020762846,
-0.0165364109,
0.0542711727,
0.0617490448,
0.0403505489,
-0.0354858786,
0.0214886554,
0.0076718368,
-0.0293594077,
-0.0609603114,
0.0018791523,
0.0049055391,
0.0307233725,
0.0339843296,
-0.068972975,
0.0244075749,
-0.0509691797,
0.001078004,
-0.01544... |
http://arxiv.org/abs/2407.01490v1 | http://arxiv.org/pdf/2407.01490v1 | LLM See, LLM Do: Guiding Data Generation to Target Non-Differentiable Objectives | 2024-07-01T17:26:21 | 2024-07-01T17:26:21 | Luísa Shimabucoro, Sebastian Ruder, Julia Kreutzer, Marzieh Fadaee, Sara Hooker | cs.CL | cs.CL, cs.AI, cs.LG | The widespread adoption of synthetic data raises new questions about how
models generating the data can influence other large language models (LLMs) via
distilled data. To start, our work exhaustively characterizes the impact of
passive inheritance of model properties by systematically studying the
consequences of synt... | LLM See, LLM Do: Guiding Data Generation to Target Non-Differentiable Objectives: The widespread adoption of synthetic data raises new questions about how
models generating the data can influence other large language models (LLMs) via
distilled data. To start, our work exhaustively characterizes the impact of
passive ... | [
0.031628862,
0.0873513147,
-0.051858101,
0.0353816263,
-0.0307855029,
0.0065018511,
0.0020512177,
0.0084651671,
-0.0081162751,
-0.0528455079,
-0.0178968944,
-0.0506951809,
0.017361952,
0.0311570633,
0.0432358161,
-0.0060589123,
0.0760068521,
-0.0230996087,
-0.0264856797,
-0.022... |
http://arxiv.org/abs/2407.01489v1 | http://arxiv.org/pdf/2407.01489v1 | Agentless: Demystifying LLM-based Software Engineering Agents | 2024-07-01T17:24:45 | 2024-07-01T17:24:45 | Chunqiu Steven Xia, Yinlin Deng, Soren Dunn, Lingming Zhang | cs.SE | cs.SE, cs.AI, cs.CL, cs.LG | Recent advancements in large language models (LLMs) have significantly
advanced the automation of software development tasks, including code
synthesis, program repair, and test generation. More recently, researchers and
industry practitioners have developed various autonomous LLM agents to perform
end-to-end software d... | Agentless: Demystifying LLM-based Software Engineering Agents: Recent advancements in large language models (LLMs) have significantly
advanced the automation of software development tasks, including code
synthesis, program repair, and test generation. More recently, researchers and
industry practitioners have developed... | [
-0.003866541,
0.0306631476,
-0.0472355224,
0.0073810392,
-0.0361197926,
-0.00413574,
-0.0124027524,
0.0162990391,
0.0109458156,
-0.0443345085,
0.0357044414,
0.049824059,
-0.0270922165,
0.0288854372,
0.016302878,
-0.0255867038,
0.0485940166,
-0.0368958861,
-0.0267808307,
0.02573... |
http://arxiv.org/abs/2407.01476v1 | http://arxiv.org/pdf/2407.01476v1 | Tree Search for Language Model Agents | 2024-07-01T17:07:55 | 2024-07-01T17:07:55 | Jing Yu Koh, Stephen McAleer, Daniel Fried, Ruslan Salakhutdinov | cs.AI | cs.AI, cs.CL, cs.LG | Autonomous agents powered by language models (LMs) have demonstrated promise
in their ability to perform decision-making tasks such as web automation.
However, a key limitation remains: LMs, primarily optimized for natural
language understanding and generation, struggle with multi-step reasoning,
planning, and using en... | Tree Search for Language Model Agents: Autonomous agents powered by language models (LMs) have demonstrated promise
in their ability to perform decision-making tasks such as web automation.
However, a key limitation remains: LMs, primarily optimized for natural
language understanding and generation, struggle with multi... | [
0.0618941784,
0.0207390767,
-0.0275412146,
0.0135918772,
-0.0507965796,
-0.0161261,
-0.0209954977,
0.013359555,
0.036251571,
-0.029917907,
-0.0186734255,
-0.0272015799,
-0.0339131169,
0.0511703081,
-0.0286964905,
-0.0589774735,
0.0125103891,
-0.0277532469,
0.038293343,
-0.00409... |
http://arxiv.org/abs/2407.01470v1 | http://arxiv.org/pdf/2407.01470v1 | DogeRM: Equipping Reward Models with Domain Knowledge through Model Merging | 2024-07-01T17:01:54 | 2024-07-01T17:01:54 | Tzu-Han Lin, Chen-An Li, Hung-yi Lee, Yun-Nung Chen | cs.CL | cs.CL | Reinforcement learning from human feedback (RLHF) is a popular strategy for
aligning large language models (LLMs) with desired behaviors. Reward modeling
is a crucial step in RLHF. However, collecting paired preference data for
training reward models is often costly and time-consuming, especially for
domain-specific pr... | DogeRM: Equipping Reward Models with Domain Knowledge through Model Merging: Reinforcement learning from human feedback (RLHF) is a popular strategy for
aligning large language models (LLMs) with desired behaviors. Reward modeling
is a crucial step in RLHF. However, collecting paired preference data for
training rewar... | [
0.0287590921,
0.0666093975,
0.0022347441,
0.0152694657,
-0.033089187,
0.0295357592,
0.0065153893,
-0.0330444463,
-0.0272813793,
-0.0273173358,
-0.0849629268,
0.0094758868,
0.0216482952,
0.0199244265,
0.0221279208,
-0.0602858551,
0.0481648967,
-0.0523052588,
0.0120495073,
-0.014... |
http://arxiv.org/abs/2407.01463v1 | http://arxiv.org/pdf/2407.01463v1 | Retrieval-augmented generation in multilingual settings | 2024-07-01T16:56:50 | 2024-07-01T16:56:50 | Nadezhda Chirkova, David Rau, Hervé Déjean, Thibault Formal, Stéphane Clinchant, Vassilina Nikoulina | cs.CL | cs.CL, cs.AI | Retrieval-augmented generation (RAG) has recently emerged as a promising
solution for incorporating up-to-date or domain-specific knowledge into large
language models (LLMs) and improving LLM factuality, but is predominantly
studied in English-only settings. In this work, we consider RAG in the
multilingual setting (mR... | Retrieval-augmented generation in multilingual settings: Retrieval-augmented generation (RAG) has recently emerged as a promising
solution for incorporating up-to-date or domain-specific knowledge into large
language models (LLMs) and improving LLM factuality, but is predominantly
studied in English-only settings. In t... | [
0.0453220122,
0.032641083,
-0.0104489438,
0.0409822613,
-0.0363571644,
-0.0177599266,
0.0106221428,
-0.0190276355,
0.0179301482,
-0.0255083814,
-0.0124005591,
-0.0536536053,
-0.0244087707,
0.0089063458,
0.0454987921,
-0.0152165256,
0.0509369895,
-0.0047918353,
-0.0422368906,
-0... |
http://arxiv.org/abs/2407.01461v1 | http://arxiv.org/pdf/2407.01461v1 | Enhancing the Capability and Robustness of Large Language Models through Reinforcement Learning-Driven Query Refinement | 2024-07-01T16:55:28 | 2024-07-01T16:55:28 | Zisu Huang, Xiaohua Wang, Feiran Zhang, Zhibo Xu, Cenyuan Zhang, Xiaoqing Zheng, Xuanjing Huang | cs.CL | cs.CL | The capacity of large language models (LLMs) to generate honest, harmless,
and helpful responses heavily relies on the quality of user prompts. However,
these prompts often tend to be brief and vague, thereby significantly limiting
the full potential of LLMs. Moreover, harmful prompts can be meticulously
crafted and ma... | Enhancing the Capability and Robustness of Large Language Models through Reinforcement Learning-Driven Query Refinement: The capacity of large language models (LLMs) to generate honest, harmless,
and helpful responses heavily relies on the quality of user prompts. However,
these prompts often tend to be brief and vagu... | [
0.0796767995,
0.0805884302,
-0.0219128076,
0.0198071059,
-0.0239061657,
-0.0169570204,
-0.0151996817,
0.0096954219,
0.01843936,
-0.0280118082,
-0.0178339574,
-0.0223726667,
-0.0116509357,
0.0130452551,
0.016701743,
-0.0434439406,
0.0634304658,
-0.0461513139,
0.0149011537,
0.006... |
http://arxiv.org/abs/2407.01455v1 | http://arxiv.org/pdf/2407.01455v1 | TimeToM: Temporal Space is the Key to Unlocking the Door of Large Language Models' Theory-of-Mind | 2024-07-01T16:50:49 | 2024-07-01T16:50:49 | Guiyang Hou, Wenqi Zhang, Yongliang Shen, Linjuan Wu, Weiming Lu | cs.CL | cs.CL | Theory of Mind (ToM)-the cognitive ability to reason about mental states of
ourselves and others, is the foundation of social interaction. Although ToM
comes naturally to humans, it poses a significant challenge to even the most
advanced Large Language Models (LLMs). Due to the complex logical chains in ToM
reasoning, ... | TimeToM: Temporal Space is the Key to Unlocking the Door of Large Language Models' Theory-of-Mind: Theory of Mind (ToM)-the cognitive ability to reason about mental states of
ourselves and others, is the foundation of social interaction. Although ToM
comes naturally to humans, it poses a significant challenge to even ... | [
0.0528642908,
0.0527979508,
-0.0240737926,
0.0482651815,
-0.0593670644,
-0.0036835105,
0.0011799483,
-0.0113767721,
0.0643209517,
-0.0142357517,
-0.0117868669,
-0.0550149605,
0.0269204993,
0.0292790662,
0.0110905049,
-0.058142975,
0.0344384611,
-0.0266836397,
0.009926849,
-0.01... |
http://arxiv.org/abs/2407.01437v1 | http://arxiv.org/pdf/2407.01437v1 | Needle in the Haystack for Memory Based Large Language Models | 2024-07-01T16:32:16 | 2024-07-01T16:32:16 | Subhajit Chaudhury, Soham Dan, Payel Das, Georgios Kollias, Elliot Nelson | cs.CL | cs.CL, cs.AI, cs.LG | In this paper, we demonstrate the benefits of using memory augmented Large
Language Model (LLM) architecture in improving the recall abilities of facts
from a potentially long context. As a case study we test LARIMAR, a recently
proposed LLM architecture which augments a LLM decoder with an external
associative memory,... | Needle in the Haystack for Memory Based Large Language Models: In this paper, we demonstrate the benefits of using memory augmented Large
Language Model (LLM) architecture in improving the recall abilities of facts
from a potentially long context. As a case study we test LARIMAR, a recently
proposed LLM architecture wh... | [
0.0133804986,
0.0307767019,
0.0032053364,
0.0372070409,
-0.0212484114,
0.0292630568,
0.0105408886,
-0.0299796946,
0.0076037962,
-0.0639222488,
-0.002533708,
-0.055003535,
0.0134745492,
0.0156048359,
0.0120235495,
-0.0236793384,
0.048120074,
-0.0705223456,
0.0083491504,
-0.00441... |
http://arxiv.org/abs/2407.01424v1 | http://arxiv.org/pdf/2407.01424v1 | A Global-Local Attention Mechanism for Relation Classification | 2024-07-01T16:14:25 | 2024-07-01T16:14:25 | Yiping Sun | cs.CL | cs.CL, cs.IR | Relation classification, a crucial component of relation extraction, involves
identifying connections between two entities. Previous studies have
predominantly focused on integrating the attention mechanism into relation
classification at a global scale, overlooking the importance of the local
context. To address this ... | A Global-Local Attention Mechanism for Relation Classification: Relation classification, a crucial component of relation extraction, involves
identifying connections between two entities. Previous studies have
predominantly focused on integrating the attention mechanism into relation
classification at a global scale, o... | [
0.0525465421,
0.0083851907,
-0.0056644566,
0.070257917,
0.006348032,
0.009008592,
-0.0001568856,
-0.0026387128,
0.0205180161,
-0.0162951183,
-0.0266838185,
-0.051204171,
-0.0043190992,
-0.0035703618,
0.0372983925,
-0.0583302118,
0.0558386408,
0.0531948879,
0.0297995042,
0.02527... |
http://arxiv.org/abs/2407.01411v2 | http://arxiv.org/pdf/2407.01411v2 | HyperLoader: Integrating Hypernetwork-Based LoRA and Adapter Layers into Multi-Task Transformers for Sequence Labelling | 2024-07-01T16:00:53 | 2024-07-02T06:21:41 | Jesus-German Ortiz-Barajas, Helena Gomez-Adorno, Thamar Solorio | cs.CL | cs.CL | We present HyperLoader, a simple approach that combines different
parameter-efficient fine-tuning methods in a multi-task setting. To achieve
this goal, our model uses a hypernetwork to generate the weights of these
modules based on the task, the transformer layer, and its position within this
layer. Our method combine... | HyperLoader: Integrating Hypernetwork-Based LoRA and Adapter Layers into Multi-Task Transformers for Sequence Labelling: We present HyperLoader, a simple approach that combines different
parameter-efficient fine-tuning methods in a multi-task setting. To achieve
this goal, our model uses a hypernetwork to generate the... | [
0.0286414418,
0.0203713123,
0.0032116605,
-0.0262140967,
-0.0085183121,
0.0645648763,
0.0677751079,
-0.0260564741,
0.0054688402,
-0.007961696,
-0.0573952794,
-0.0231877584,
0.0066778352,
-0.0022407924,
0.0329621769,
0.0210594628,
0.0174404513,
0.0133044571,
0.0125762168,
0.0262... |
http://arxiv.org/abs/2407.01409v1 | http://arxiv.org/pdf/2407.01409v1 | Dynamic Few-Shot Learning for Knowledge Graph Question Answering | 2024-07-01T15:59:17 | 2024-07-01T15:59:17 | Jacopo D'Abramo, Andrea Zugarini, Paolo Torroni | cs.CL | cs.CL, cs.AI | Large language models present opportunities for innovative Question Answering
over Knowledge Graphs (KGQA). However, they are not inherently designed for
query generation. To bridge this gap, solutions have been proposed that rely on
fine-tuning or ad-hoc architectures, achieving good results but limited
out-of-domain ... | Dynamic Few-Shot Learning for Knowledge Graph Question Answering: Large language models present opportunities for innovative Question Answering
over Knowledge Graphs (KGQA). However, they are not inherently designed for
query generation. To bridge this gap, solutions have been proposed that rely on
fine-tuning or ad-ho... | [
0.0394126922,
0.0031926122,
0.0086710174,
0.015945131,
-0.0382769108,
0.0143420463,
-0.011722791,
-0.0255802162,
-0.0076578241,
-0.0550526455,
-0.0139734671,
-0.009137651,
-0.0420125127,
0.0102689471,
0.0234537069,
0.0133634508,
0.0247671586,
0.0171076395,
0.0068334467,
-0.0152... |
http://arxiv.org/abs/2407.01406v1 | http://arxiv.org/pdf/2407.01406v1 | Adapting Multilingual LLMs to Low-Resource Languages with Knowledge Graphs via Adapters | 2024-07-01T15:56:24 | 2024-07-01T15:56:24 | Daniil Gurgurov, Mareike Hartmann, Simon Ostermann | cs.CL | cs.CL, cs.AI | This paper explores the integration of graph knowledge from linguistic
ontologies into multilingual Large Language Models (LLMs) using adapters to
improve performance for low-resource languages (LRLs) in sentiment analysis
(SA) and named entity recognition (NER). Building upon successful
parameter-efficient fine-tuning... | Adapting Multilingual LLMs to Low-Resource Languages with Knowledge Graphs via Adapters: This paper explores the integration of graph knowledge from linguistic
ontologies into multilingual Large Language Models (LLMs) using adapters to
improve performance for low-resource languages (LRLs) in sentiment analysis
(SA) an... | [
0.0422703736,
0.0329497047,
-0.0068400134,
0.0812963992,
-0.0316445939,
0.0340649672,
-0.0171665568,
0.0121488404,
0.0553599969,
-0.0246551782,
-0.0318910219,
-0.0096079176,
-0.0063967234,
0.0022223548,
0.010922594,
-0.0218070131,
0.0221201517,
-0.0240743682,
0.0251356307,
-0.0... |
http://arxiv.org/abs/2407.01403v1 | http://arxiv.org/pdf/2407.01403v1 | Optimization of Retrieval-Augmented Generation Context with Outlier Detection | 2024-07-01T15:53:29 | 2024-07-01T15:53:29 | Vitaly Bulgakov | cs.IR | cs.IR, cs.AI, cs.CL, cs.LG | In this paper, we focus on methods to reduce the size and improve the quality
of the prompt context required for question-answering systems. Attempts to
increase the number of retrieved chunked documents and thereby enlarge the
context related to the query can significantly complicate the processing and
decrease the pe... | Optimization of Retrieval-Augmented Generation Context with Outlier Detection: In this paper, we focus on methods to reduce the size and improve the quality
of the prompt context required for question-answering systems. Attempts to
increase the number of retrieved chunked documents and thereby enlarge the
context rela... | [
0.0579806343,
-0.012858009,
-0.0152273215,
0.0468550883,
-0.0328062028,
0.0135269025,
-0.0147611778,
0.0245893281,
-0.0177726466,
-0.0330129415,
0.0281904675,
-0.0401042551,
-0.0387083367,
-0.0105165262,
-0.0106737372,
-0.0216253716,
0.0382708907,
-0.0327280313,
0.0375985689,
-... |
http://arxiv.org/abs/2407.01394v1 | http://arxiv.org/pdf/2407.01394v1 | Gloss2Text: Sign Language Gloss translation using LLMs and Semantically Aware Label Smoothing | 2024-07-01T15:46:45 | 2024-07-01T15:46:45 | Pooya Fayyazsanavi, Antonios Anastasopoulos, Jana Košecká | cs.CV | cs.CV, cs.CL, cs.LG | Sign language translation from video to spoken text presents unique
challenges owing to the distinct grammar, expression nuances, and high
variation of visual appearance across different speakers and contexts. The
intermediate gloss annotations of videos aim to guide the translation process.
In our work, we focus on {\... | Gloss2Text: Sign Language Gloss translation using LLMs and Semantically Aware Label Smoothing: Sign language translation from video to spoken text presents unique
challenges owing to the distinct grammar, expression nuances, and high
variation of visual appearance across different speakers and contexts. The
intermedia... | [
-0.0336639099,
0.0407453477,
0.0054755732,
0.0786297694,
0.0169876199,
0.034477625,
0.0065078065,
0.0038376749,
0.0363627784,
-0.0407052301,
-0.0310601909,
-0.0412637107,
0.0275926348,
-0.0016330226,
0.0207810905,
-0.1017910987,
0.0347370729,
-0.0375723839,
-0.0320760868,
-0.01... |
http://arxiv.org/abs/2407.01393v1 | http://arxiv.org/pdf/2407.01393v1 | POLygraph: Polish Fake News Dataset | 2024-07-01T15:45:21 | 2024-07-01T15:45:21 | Daniel Dzienisiewicz, Filip Graliński, Piotr Jabłoński, Marek Kubis, Paweł Skórzewski, Piotr Wierzchoń | cs.CL | cs.CL | This paper presents the POLygraph dataset, a unique resource for fake news
detection in Polish. The dataset, created by an interdisciplinary team, is
composed of two parts: the "fake-or-not" dataset with 11,360 pairs of news
articles (identified by their URLs) and corresponding labels, and the
"fake-they-say" dataset w... | POLygraph: Polish Fake News Dataset: This paper presents the POLygraph dataset, a unique resource for fake news
detection in Polish. The dataset, created by an interdisciplinary team, is
composed of two parts: the "fake-or-not" dataset with 11,360 pairs of news
articles (identified by their URLs) and corresponding labe... | [
0.0811020806,
0.0332393013,
-0.0100592542,
0.050640244,
-0.0499266163,
0.0391588509,
0.0276436191,
-0.0036768019,
0.0199757237,
-0.003341231,
0.0452033654,
0.0323111154,
0.0138990507,
0.068226479,
0.013170965,
-0.00941859,
0.0542018265,
0.0429387689,
0.0693856031,
-0.0207811948... |
http://arxiv.org/abs/2407.01384v1 | http://arxiv.org/pdf/2407.01384v1 | Free-text Rationale Generation under Readability Level Control | 2024-07-01T15:34:17 | 2024-07-01T15:34:17 | Yi-Sheng Hsu, Nils Feldhus, Sherzod Hakimov | cs.CL | cs.CL | Free-text rationales justify model decisions in natural language and thus
become likable and accessible among approaches to explanation across many
tasks. However, their effectiveness can be hindered by misinterpretation and
hallucination. As a perturbation test, we investigate how large language models
(LLMs) perform ... | Free-text Rationale Generation under Readability Level Control: Free-text rationales justify model decisions in natural language and thus
become likable and accessible among approaches to explanation across many
tasks. However, their effectiveness can be hindered by misinterpretation and
hallucination. As a perturbatio... | [
0.0183457416,
0.0255935676,
-0.0068051945,
0.0174803678,
-0.0248013511,
0.0195535738,
0.0537705533,
-0.0195072349,
0.004155281,
-0.0249016881,
0.0363760851,
-0.0390475914,
0.0314580128,
0.0123085082,
0.0192618538,
-0.0703272521,
0.0029945679,
-0.0411961377,
0.063835822,
0.00429... |
http://arxiv.org/abs/2407.01376v1 | http://arxiv.org/pdf/2407.01376v1 | Badllama 3: removing safety finetuning from Llama 3 in minutes | 2024-07-01T15:29:45 | 2024-07-01T15:29:45 | Dmitrii Volkov | cs.LG | cs.LG, cs.AI, cs.CL, cs.CR | We show that extensive LLM safety fine-tuning is easily subverted when an
attacker has access to model weights. We evaluate three state-of-the-art
fine-tuning methods-QLoRA, ReFT, and Ortho-and show how algorithmic advances
enable constant jailbreaking performance with cuts in FLOPs and optimisation
power. We strip saf... | Badllama 3: removing safety finetuning from Llama 3 in minutes: We show that extensive LLM safety fine-tuning is easily subverted when an
attacker has access to model weights. We evaluate three state-of-the-art
fine-tuning methods-QLoRA, ReFT, and Ortho-and show how algorithmic advances
enable constant jailbreaking per... | [
-0.0057324572,
0.0193817485,
-0.0134619242,
0.056835752,
-0.0205029175,
-0.0170996413,
-0.0036836946,
-0.0305209029,
0.0070714052,
-0.0041230815,
-0.0176325329,
0.0174586345,
0.009400255,
0.0066693719,
0.0406143889,
0.0444137529,
0.0642563701,
-0.0243054982,
-0.0220500622,
0.01... |
http://arxiv.org/abs/2407.01374v1 | http://arxiv.org/pdf/2407.01374v1 | Bridging the Gap: Transfer Learning from English PLMs to Malaysian English | 2024-07-01T15:26:03 | 2024-07-01T15:26:03 | Mohan Raj Chanthran, Lay-Ki Soon, Huey Fang Ong, Bhawani Selvaretnam | cs.CL | cs.CL | Malaysian English is a low resource creole language, where it carries the
elements of Malay, Chinese, and Tamil languages, in addition to Standard
English. Named Entity Recognition (NER) models underperform when capturing
entities from Malaysian English text due to its distinctive morphosyntactic
adaptations, semantic ... | Bridging the Gap: Transfer Learning from English PLMs to Malaysian English: Malaysian English is a low resource creole language, where it carries the
elements of Malay, Chinese, and Tamil languages, in addition to Standard
English. Named Entity Recognition (NER) models underperform when capturing
entities from Malaysi... | [
0.0337770618,
0.0037592449,
0.0112896375,
0.0316050164,
-0.0086183306,
0.0459962524,
-0.0096039502,
-0.0156185692,
-0.0242289286,
-0.05337964,
-0.0660872385,
-0.046499189,
0.0043816157,
-0.0355481431,
0.0380874276,
-0.0647416711,
0.0220007282,
-0.0074704108,
0.032720577,
0.0064... |
http://arxiv.org/abs/2407.01370v1 | http://arxiv.org/pdf/2407.01370v1 | Summary of a Haystack: A Challenge to Long-Context LLMs and RAG Systems | 2024-07-01T15:23:42 | 2024-07-01T15:23:42 | Philippe Laban, Alexander R. Fabbri, Caiming Xiong, Chien-Sheng Wu | cs.CL | cs.CL | LLMs and RAG systems are now capable of handling millions of input tokens or
more. However, evaluating the output quality of such systems on long-context
tasks remains challenging, as tasks like Needle-in-a-Haystack lack complexity.
In this work, we argue that summarization can play a central role in such
evaluation. W... | Summary of a Haystack: A Challenge to Long-Context LLMs and RAG Systems: LLMs and RAG systems are now capable of handling millions of input tokens or
more. However, evaluating the output quality of such systems on long-context
tasks remains challenging, as tasks like Needle-in-a-Haystack lack complexity.
In this work, ... | [
0.0522320867,
-0.0226815455,
0.0085100578,
-0.0018591688,
-0.0433894284,
0.0000207944,
-0.0223783422,
0.0072510149,
-0.0068129068,
-0.072583504,
-0.0219879001,
0.010048762,
-0.0046318015,
-0.0281048883,
0.0218146723,
-0.0211170074,
0.0437504575,
-0.004147999,
-0.0184954163,
-0.... |
http://arxiv.org/abs/2407.01360v1 | http://arxiv.org/pdf/2407.01360v1 | Nullpointer at ArAIEval Shared Task: Arabic Propagandist Technique Detection with Token-to-Word Mapping in Sequence Tagging | 2024-07-01T15:15:24 | 2024-07-01T15:15:24 | Abrar Abir, Kemal Oflazer | cs.CL | cs.CL | This paper investigates the optimization of propaganda technique detection in
Arabic text, including tweets \& news paragraphs, from ArAIEval shared task 1.
Our approach involves fine-tuning the AraBERT v2 model with a neural network
classifier for sequence tagging. Experimental results show relying on the first
token ... | Nullpointer at ArAIEval Shared Task: Arabic Propagandist Technique Detection with Token-to-Word Mapping in Sequence Tagging: This paper investigates the optimization of propaganda technique detection in
Arabic text, including tweets \& news paragraphs, from ArAIEval shared task 1.
Our approach involves fine-tuning the... | [
0.0455777831,
0.0183543134,
-0.0141829457,
-0.0028459828,
-0.0559174046,
0.0082960064,
0.0324498825,
0.0270865411,
-0.0473381318,
-0.0299007576,
0.0170326494,
-0.0670234635,
0.0577749945,
0.0421226509,
0.0683820397,
0.0192961674,
0.0439134613,
0.0015516941,
0.050086841,
-0.0055... |
http://arxiv.org/abs/2407.01358v1 | http://arxiv.org/pdf/2407.01358v1 | Evaluating Knowledge-based Cross-lingual Inconsistency in Large Language Models | 2024-07-01T15:11:37 | 2024-07-01T15:11:37 | Xiaolin Xing, Zhiwei He, Haoyu Xu, Xing Wang, Rui Wang, Yu Hong | cs.CL | cs.CL | This paper investigates the cross-lingual inconsistencies observed in Large
Language Models (LLMs), such as ChatGPT, Llama, and Baichuan, which have shown
exceptional performance in various Natural Language Processing (NLP) tasks.
Despite their successes, these models often exhibit significant inconsistencies
when proc... | Evaluating Knowledge-based Cross-lingual Inconsistency in Large Language Models: This paper investigates the cross-lingual inconsistencies observed in Large
Language Models (LLMs), such as ChatGPT, Llama, and Baichuan, which have shown
exceptional performance in various Natural Language Processing (NLP) tasks.
Despite... | [
0.0434774794,
0.0117063131,
-0.0394027829,
0.0723245665,
-0.0726112351,
0.0029771458,
0.0040976969,
-0.0263949856,
0.0273342673,
-0.0314338803,
-0.0440474749,
-0.0595652014,
0.0504835322,
0.0037111174,
-0.0089275464,
-0.0200195126,
0.0833289772,
-0.0075868662,
-0.0294541717,
-0... |
http://arxiv.org/abs/2407.01334v2 | http://arxiv.org/pdf/2407.01334v2 | Protecting Privacy in Classifiers by Token Manipulation | 2024-07-01T14:41:59 | 2024-07-03T16:31:52 | Re'em Harel, Yair Elboher, Yuval Pinter | cs.CL | cs.CL, cs.CR | Using language models as a remote service entails sending private information
to an untrusted provider. In addition, potential eavesdroppers can intercept
the messages, thereby exposing the information. In this work, we explore the
prospects of avoiding such data exposure at the level of text manipulation. We
focus on ... | Protecting Privacy in Classifiers by Token Manipulation: Using language models as a remote service entails sending private information
to an untrusted provider. In addition, potential eavesdroppers can intercept
the messages, thereby exposing the information. In this work, we explore the
prospects of avoiding such data... | [
0.0350912176,
0.0693323761,
-0.0149090495,
0.0439432338,
-0.0964030027,
-0.0102598323,
0.0782079175,
-0.0020722169,
-0.0008515119,
-0.0728632212,
-0.0221867282,
0.0059361481,
0.0216521192,
0.0501036383,
0.0162634254,
0.0099537857,
0.0128387222,
-0.031467218,
-0.0175378285,
-0.0... |
http://arxiv.org/abs/2407.01320v1 | http://arxiv.org/pdf/2407.01320v1 | Increasing Model Capacity for Free: A Simple Strategy for Parameter Efficient Fine-tuning | 2024-07-01T14:26:48 | 2024-07-01T14:26:48 | Haobo Song, Hao Zhao, Soumajit Majumder, Tao Lin | cs.LG | cs.LG, cs.AI, cs.CL | Fine-tuning large pre-trained foundation models, such as the 175B GPT-3, has
attracted more attention for downstream tasks recently. While
parameter-efficient fine-tuning methods have been proposed and proven effective
without retraining all model parameters, their performance is limited by the
capacity of incremental ... | Increasing Model Capacity for Free: A Simple Strategy for Parameter Efficient Fine-tuning: Fine-tuning large pre-trained foundation models, such as the 175B GPT-3, has
attracted more attention for downstream tasks recently. While
parameter-efficient fine-tuning methods have been proposed and proven effective
without r... | [
0.0354469903,
0.066122897,
-0.0021964407,
0.0354526974,
0.0290401522,
0.0158341844,
0.0123814894,
-0.0235840026,
-0.0046366425,
-0.0288162157,
-0.0459123738,
-0.0216900706,
-0.0104574952,
0.0307383407,
0.0183939002,
-0.060589347,
0.0378283598,
-0.0017924539,
-0.0358548649,
0.00... |
http://arxiv.org/abs/2407.01315v1 | http://arxiv.org/pdf/2407.01315v1 | Language Portability Strategies for Open-domain Dialogue with Pre-trained Language Models from High to Low Resource Languages | 2024-07-01T14:20:54 | 2024-07-01T14:20:54 | Ahmed Njifenjou, Virgile Sucal, Bassam Jabaian, Fabrice Lefèvre | cs.CL | cs.CL | In this paper we propose a study of linguistic portability strategies of
large pre-trained language models (PLMs) used for open-domain dialogue systems
in a high-resource language for this task. In particular the target
low-resource language (L_T) will be simulated with French, as it lacks of
task-specific resources an... | Language Portability Strategies for Open-domain Dialogue with Pre-trained Language Models from High to Low Resource Languages: In this paper we propose a study of linguistic portability strategies of
large pre-trained language models (PLMs) used for open-domain dialogue systems
in a high-resource language for this tas... | [
0.0011349907,
0.0625068992,
-0.0175477136,
0.068297036,
-0.0330315121,
-0.0106679639,
0.0375414304,
-0.0109034861,
-0.0046887919,
-0.0237192754,
-0.0955831483,
-0.0304415263,
-0.0149013074,
0.0326076224,
0.0086002657,
-0.0243498851,
0.0535938405,
-0.0363699161,
-0.0341224,
0.00... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.