id stringlengths 33 34 | doc_url stringlengths 33 34 | title stringlengths 10 211 | publication_date timestamp[s] | update_date timestamp[s] | authors stringlengths 5 20.1k | category_primary stringclasses 78
values | category_all stringlengths 5 138 | abstract stringlengths 123 3.09k | embeddings listlengths 768 768 |
|---|---|---|---|---|---|---|---|---|---|
http://arxiv.org/abs/2406.15796v1 | http://arxiv.org/pdf/2406.15796v1 | Rethinking Entity-level Unlearning for Large Language Models | 2024-06-22T09:40:07 | 2024-06-22T09:40:07 | Weitao Ma, Xiaocheng Feng, Weihong Zhong, Lei Huang, Yangfan Ye, Bing Qin | cs.CL | cs.CL | Large language model unlearning has gained increasing attention due to its
potential to mitigate security and privacy concerns. Current research
predominantly focuses on Instance-level unlearning, specifically aiming at
forgetting predefined instances of sensitive content. However, a notable gap
still exists in explori... | [
-0.425927788,
-0.8997608423,
0.4765038192,
-0.5415128469,
-0.4709865451,
0.2076087594,
-0.4838911891,
-0.5143863559,
-0.2598201036,
0.6832871437,
-0.4827682972,
-0.593042016,
-0.3138386607,
-0.3439656198,
-0.6626887321,
1.2077417374,
0.0658467487,
-0.1745356172,
-0.2983722687,
... |
http://arxiv.org/abs/2406.15786v1 | http://arxiv.org/pdf/2406.15786v1 | What Matters in Transformers? Not All Attention is Needed | 2024-06-22T08:41:48 | 2024-06-22T08:41:48 | Shwai He, Guoheng Sun, Zheyu Shen, Ang Li | cs.LG | cs.LG, cs.AI, cs.CL | Scaling Transformer-based large language models (LLMs) has demonstrated
promising performance across various tasks. However, this scaling also
introduces redundant structures, posing challenges for real-world deployment.
Despite some recognition of redundancy in LLMs, the variability of redundancy
across different stru... | [
-0.7850154042,
-0.3159024417,
0.1661776602,
0.474622786,
0.1524161249,
-0.1177253872,
-0.1750682145,
-0.5440582037,
0.1624432355,
0.4898008406,
-0.5942907929,
-0.0548351891,
-0.7077963948,
0.1674057394,
-0.7048581839,
0.9307411313,
-0.0141678834,
-0.4139658809,
-0.4532862604,
-... |
http://arxiv.org/abs/2406.15781v1 | http://arxiv.org/pdf/2406.15781v1 | DABL: Detecting Semantic Anomalies in Business Processes Using Large Language Models | 2024-06-22T08:20:19 | 2024-06-22T08:20:19 | Wei Guan, Jian Cao, Jianqi Gao, Haiyan Zhao, Shiyou Qian | cs.CL | cs.CL | Detecting anomalies in business processes is crucial for ensuring operational
success. While many existing methods rely on statistical frequency to detect
anomalies, it's important to note that infrequent behavior doesn't necessarily
imply undesirability. To address this challenge, detecting anomalies from a
semantic v... | [
-0.1712168902,
-1.3304519653,
0.7716326714,
0.0463068038,
-0.203665942,
-0.0195559561,
0.1450508386,
-0.7721195817,
0.2159295082,
0.8914300203,
-0.9360553026,
-0.6233417988,
-0.7066013813,
-0.0711336657,
-0.6432837844,
0.933989346,
0.0692626014,
-0.4102721214,
-0.0116003584,
-0... |
http://arxiv.org/abs/2406.15765v1 | http://arxiv.org/pdf/2406.15765v1 | Unveiling and Harnessing Hidden Attention Sinks: Enhancing Large Language Models without Training through Attention Calibration | 2024-06-22T07:00:43 | 2024-06-22T07:00:43 | Zhongzhi Yu, Zheng Wang, Yonggan Fu, Huihong Shi, Khalid Shaikh, Yingyan Celine Lin | cs.LG | cs.LG, cs.CL | Attention is a fundamental component behind the remarkable achievements of
large language models (LLMs). However, our current understanding of the
attention mechanism, especially regarding how attention distributions are
established, remains limited. Inspired by recent studies that explore the
presence of attention sin... | [
-0.4784597754,
-0.4943520427,
0.8820870519,
0.1181993559,
-0.2956478894,
0.1683788151,
-0.2606239915,
-0.1139862016,
0.2822450101,
0.5948262811,
-0.6332389712,
-0.3377917707,
-0.7917414904,
-0.0564990975,
-0.3135997951,
0.7471422553,
0.1257272363,
-0.0318802483,
-0.3175336123,
... |
http://arxiv.org/abs/2406.15754v1 | http://arxiv.org/pdf/2406.15754v1 | Multimodal Segmentation for Vocal Tract Modeling | 2024-06-22T06:44:38 | 2024-06-22T06:44:38 | Rishi Jain, Bohan Yu, Peter Wu, Tejas Prabhune, Gopala Anumanchipalli | cs.CV | cs.CV, cs.CL, cs.LG, cs.SD, eess.AS | Accurate modeling of the vocal tract is necessary to construct articulatory
representations for interpretable speech processing and linguistics. However,
vocal tract modeling is challenging because many internal articulators are
occluded from external motion capture technologies. Real-time magnetic
resonance imaging (R... | [
-0.5304631591,
-0.5806987286,
0.3218982816,
-0.0544860847,
-0.2630609572,
0.1863172501,
-0.2204017639,
-0.4296103716,
0.2996530235,
0.4338406622,
-0.8724194765,
-0.3521624506,
-0.4921172559,
-0.1738146693,
-0.0539236777,
0.6646162271,
0.3093639314,
0.4902198017,
-0.1405559182,
... |
http://arxiv.org/abs/2406.15752v1 | http://arxiv.org/pdf/2406.15752v1 | TacoLM: GaTed Attention Equipped Codec Language Model are Efficient Zero-Shot Text to Speech Synthesizers | 2024-06-22T06:39:52 | 2024-06-22T06:39:52 | Yakun Song, Zhuo Chen, Xiaofei Wang, Ziyang Ma, Guanrou Yang, Xie Chen | eess.AS | eess.AS, cs.AI, cs.CL | Neural codec language model (LM) has demonstrated strong capability in
zero-shot text-to-speech (TTS) synthesis. However, the codec LM often suffers
from limitations in inference speed and stability, due to its auto-regressive
nature and implicit alignment between text and audio. In this work, to handle
these challenge... | [
-0.5235083103,
-0.7610265613,
0.3256709874,
0.3258372545,
-0.3869223893,
0.2428643107,
-0.2871613801,
-0.2304107398,
0.1920334846,
0.5122386813,
-0.3257266879,
-0.9691208005,
-0.4060188234,
-0.0186725538,
-0.3858414888,
1.0356985331,
0.0101670558,
0.0941152126,
0.1648787558,
-0... |
http://arxiv.org/abs/2406.15741v1 | http://arxiv.org/pdf/2406.15741v1 | Ladder: A Model-Agnostic Framework Boosting LLM-based Machine Translation to the Next Level | 2024-06-22T05:33:35 | 2024-06-22T05:33:35 | Zhaopeng Feng, Ruizhe Chen, Yan Zhang, Zijie Meng, Zuozhu Liu | cs.CL | cs.CL, cs.AI, cs.LG | General-purpose Large Language Models (LLMs) like GPT-4 have achieved
remarkable advancements in machine translation (MT) by leveraging extensive web
content. On the other hand, translation-specific LLMs are built by pre-training
on domain-specific monolingual corpora and fine-tuning with human-annotated
translation da... | [
0.0428702049,
-0.460362941,
0.1220097616,
0.2549213469,
-0.3458802998,
0.4015499055,
0.0836161077,
-0.3139244616,
-0.0776810274,
0.3416514397,
-0.4122632146,
-0.6306548715,
-0.6994073987,
0.2909813523,
-0.58081007,
0.999270618,
-0.0020721247,
0.18642129,
0.3363864124,
-0.090790... |
http://arxiv.org/abs/2406.15736v1 | http://arxiv.org/pdf/2406.15736v1 | Evaluating Large Vision-and-Language Models on Children's Mathematical Olympiads | 2024-06-22T05:04:39 | 2024-06-22T05:04:39 | Anoop Cherian, Kuan-Chuan Peng, Suhas Lohit, Joanna Matthiesen, Kevin Smith, Joshua B. Tenenbaum | cs.LG | cs.LG, cs.AI, cs.CL, cs.CV | Recent years have seen a significant progress in the general-purpose problem
solving abilities of large vision and language models (LVLMs), such as ChatGPT,
Gemini, etc.; some of these breakthroughs even seem to enable AI models to
outperform human abilities in varied tasks that demand higher-order cognitive
skills. Ar... | [
-0.3507926464,
-0.6586298347,
0.6784803271,
0.1687712669,
-0.23475568,
0.1487026364,
-0.2212369889,
-0.5934550166,
-0.2808282971,
0.376935035,
-0.5540420413,
-0.1283745617,
-0.5666453838,
0.2226029634,
-0.364563942,
0.3907102942,
-0.1469791979,
0.0537455678,
-0.8604934812,
0.05... |
http://arxiv.org/abs/2406.15734v1 | http://arxiv.org/pdf/2406.15734v1 | RankAdaptor: Hierarchical Dynamic Low-Rank Adaptation for Structural Pruned LLMs | 2024-06-22T04:52:58 | 2024-06-22T04:52:58 | Changhai Zhou, Shijie Han, Shiyang Zhang, Shichao Weng, Zekai Liu, Cheng Jin | cs.CL | cs.CL, cs.AI | The efficient compression of large language models (LLMs) is becoming
increasingly popular. However, recovering the accuracy of compressed LLMs is
still a major challenge. Structural pruning with standard Low-Rank Adaptation
(LoRA) is a common technique in current LLM compression. In structural pruning,
the model archi... | [
-0.0009075541,
-0.5354790688,
0.2148747891,
0.081100218,
-0.5542664528,
0.4794499576,
-0.158347711,
-0.2646244764,
0.1446247697,
0.2265471071,
-0.442135781,
-0.250180155,
-0.7596324682,
-0.0609408282,
-0.0906610414,
1.0644294024,
0.0032842518,
-0.2919498086,
0.1051360145,
-0.13... |
http://arxiv.org/abs/2406.15723v1 | http://arxiv.org/pdf/2406.15723v1 | Acoustic Feature Mixup for Balanced Multi-aspect Pronunciation Assessment | 2024-06-22T03:56:29 | 2024-06-22T03:56:29 | Heejin Do, Wonjun Lee, Gary Geunbae Lee | cs.CL | cs.CL, cs.AI, cs.SD, eess.AS | In automated pronunciation assessment, recent emphasis progressively lies on
evaluating multiple aspects to provide enriched feedback. However, acquiring
multi-aspect-score labeled data for non-native language learners' speech poses
challenges; moreover, it often leads to score-imbalanced distributions. In this
paper, ... | [
-0.829064846,
-0.5495510101,
-0.0182346497,
0.4894760549,
-0.3950569332,
0.3226955235,
-0.1877918392,
-0.6533830166,
0.7069835067,
0.3521154225,
-0.1668159962,
-0.2684217095,
-0.3958232701,
-0.1251580119,
-0.1544008404,
0.7798606157,
0.1714344472,
0.1805087775,
-0.1655323207,
-... |
http://arxiv.org/abs/2406.15720v1 | http://arxiv.org/pdf/2406.15720v1 | Scaling Laws for Fact Memorization of Large Language Models | 2024-06-22T03:32:09 | 2024-06-22T03:32:09 | Xingyu Lu, Xiaonan Li, Qinyuan Cheng, Kai Ding, Xuanjing Huang, Xipeng Qiu | cs.CL | cs.CL | Fact knowledge memorization is crucial for Large Language Models (LLM) to
generate factual and reliable responses. However, the behaviors of LLM fact
memorization remain under-explored. In this paper, we analyze the scaling laws
for LLM's fact knowledge and LLMs' behaviors of memorizing different types of
facts. We fin... | [
-0.239861846,
-0.8710534573,
0.8850973248,
-0.0652622804,
-0.313790828,
-0.3890993595,
-0.3378073871,
-0.2136179358,
0.1911798567,
0.7315146923,
-0.3029621542,
-0.2537327111,
-0.6039041877,
-0.2225823104,
-0.2245891541,
1.3120360374,
0.0970258713,
0.0615534522,
-0.6531072259,
-... |
http://arxiv.org/abs/2406.15718v1 | http://arxiv.org/pdf/2406.15718v1 | Beyond the Turn-Based Game: Enabling Real-Time Conversations with Duplex Models | 2024-06-22T03:20:10 | 2024-06-22T03:20:10 | Xinrong Zhang, Yingfa Chen, Shengding Hu, Xu Han, Zihang Xu, Yuanwei Xu, Weilin Zhao, Maosong Sun, Zhiyuan Liu | cs.CL | cs.CL | As large language models (LLMs) increasingly permeate daily lives, there is a
growing demand for real-time interactions that mirror human conversations.
Traditional turn-based chat systems driven by LLMs prevent users from verbally
interacting with the system while it is generating responses. To overcome these
limitati... | [
-0.4481893778,
-1.1886633635,
0.6662114263,
0.6447038651,
-0.1787911206,
0.6468343735,
-0.1038670018,
-0.7687388659,
0.3831635416,
0.7934023142,
-0.8023666739,
-0.2941006124,
-0.0744230598,
-0.0802515671,
-0.4824420214,
0.6249319911,
0.1849898547,
0.1895445734,
-0.3349384964,
-... |
http://arxiv.org/abs/2406.15708v1 | http://arxiv.org/pdf/2406.15708v1 | Teach Better or Show Smarter? On Instructions and Exemplars in Automatic Prompt Optimization | 2024-06-22T02:07:10 | 2024-06-22T02:07:10 | Xingchen Wan, Ruoxi Sun, Hootan Nakhost, Sercan O. Arik | cs.CL | cs.CL, cs.AI, cs.LG | Large language models have demonstrated remarkable capabilities, but their
performance is heavily reliant on effective prompt engineering. Automatic
prompt optimization (APO) methods are designed to automate this and can be
broadly categorized into those targeting instructions (instruction
optimization, IO) vs. those t... | [
-0.3042806983,
-0.6783585548,
0.5935900211,
-0.0044570337,
-0.1431000829,
0.044725202,
-0.2769185007,
-0.4203919768,
-0.0935518667,
0.0536336415,
-0.6967579126,
-0.13106291,
-0.5138350129,
0.4425585866,
-0.1847413033,
0.7660005689,
0.0156947691,
0.075457871,
-0.6761269569,
0.09... |
http://arxiv.org/abs/2406.16962v1 | http://arxiv.org/pdf/2406.16962v1 | MetaGreen: Meta-Learning Inspired Transformer Selection for Green Semantic Communication | 2024-06-22T00:49:40 | 2024-06-22T00:49:40 | Shubhabrata Mukherjee, Cory Beard, Sejun Song | cs.LG | cs.LG, cs.AI, cs.CL | Semantic Communication can transform the way we transmit information,
prioritizing meaningful and effective content over individual symbols or bits.
This evolution promises significant benefits, including reduced latency, lower
bandwidth usage, and higher throughput compared to traditional communication.
However, the d... | [
-0.4239593446,
-0.5181675553,
0.512852788,
-0.2186397165,
-0.3709241748,
-0.2093445212,
-0.154180631,
-0.5859342813,
0.2635987699,
0.2045215368,
-0.8657794595,
-0.3854945004,
-0.571634233,
0.1650077403,
-0.0919689387,
0.8768942952,
0.1530331969,
-0.0545432046,
0.1177444682,
-0.... |
http://arxiv.org/abs/2406.15695v1 | http://arxiv.org/pdf/2406.15695v1 | SS-Bench: A Benchmark for Social Story Generation and Evaluation | 2024-06-22T00:14:48 | 2024-06-22T00:14:48 | Yi Feng, Mingyang Song, Jiaqi Wang, Mao Zheng, Liping Jing, Jian Yu | cs.CL | cs.CL | Children with Autism Spectrum Disorder (ASD) often misunderstand social
situations and struggle to participate in daily routines. Psychology experts
write Social Stories under strict constraints of structural clarity,
descriptive orientation, and situational safety to enhance their abilities in
these regimes. However, ... | [
-0.2461233735,
-0.4877853692,
0.4881676733,
0.408639282,
-0.1776904315,
0.5790093541,
-0.0869646221,
-0.5263399482,
0.4849460125,
0.2527580261,
-0.8716965914,
-0.4277458489,
-0.4945774078,
0.1955557019,
-0.5234183073,
1.0287972689,
-0.2556221485,
-0.239569068,
-0.0567176528,
-0... |
http://arxiv.org/abs/2406.15673v1 | http://arxiv.org/pdf/2406.15673v1 | Large Language Models have Intrinsic Self-Correction Ability | 2024-06-21T22:29:40 | 2024-06-21T22:29:40 | Dancheng Liu, Amir Nassereldine, Ziming Yang, Chenhui Xu, Yuting Hu, Jiajie Li, Utkarsh Kumar, Changjae Lee, Jinjun Xiong | cs.CL | cs.CL, cs.AI | Large language models (LLMs) have attracted significant attention for their
remarkable abilities in various natural language processing tasks, but they
suffer from hallucinations that will cause performance degradation. One
promising solution to improve the LLMs' performance is to ask LLMs to revise
their answer after ... | [
-0.103967905,
-0.8829093575,
0.7479277849,
0.0668787211,
0.1250384599,
0.1041694358,
-0.5155113339,
-0.3389934003,
0.0449547544,
0.6402845979,
-0.6046726704,
-0.0472579971,
-0.3573089242,
0.3678622842,
-0.4739329815,
1.0017454624,
0.0347484313,
-0.0635256022,
-0.0571831502,
-0.... |
http://arxiv.org/abs/2406.15668v1 | http://arxiv.org/pdf/2406.15668v1 | PI-Whisper: An Adaptive and Incremental ASR Framework for Diverse and Evolving Speaker Characteristics | 2024-06-21T21:58:37 | 2024-06-21T21:58:37 | Amir Nassereldine, Dancheng Liu, Chenhui Xu, Jinjun Xiong | cs.CL | cs.CL, cs.SD, eess.AS | As edge-based automatic speech recognition (ASR) technologies become
increasingly prevalent for the development of intelligent and personalized
assistants, three important challenges must be addressed for these
resource-constrained ASR models, i.e., adaptivity, incrementality, and
inclusivity. We propose a novel ASR fr... | [
-0.4686072171,
-0.5669885874,
0.0289322361,
0.5063048005,
0.3214211762,
0.4286970794,
-0.2901703119,
-0.6048673987,
0.3343218267,
0.0802030563,
-0.7182102203,
-0.1323155761,
-0.6345750093,
-0.0134358453,
-0.4370386004,
0.5111662745,
0.4674796462,
0.0868430063,
0.300316602,
-0.3... |
http://arxiv.org/abs/2406.15627v1 | http://arxiv.org/pdf/2406.15627v1 | Benchmarking Uncertainty Quantification Methods for Large Language Models with LM-Polygraph | 2024-06-21T20:06:31 | 2024-06-21T20:06:31 | Roman Vashurin, Ekaterina Fadeeva, Artem Vazhentsev, Akim Tsvigun, Daniil Vasilev, Rui Xing, Abdelrahman Boda Sadallah, Lyudmila Rvanova, Sergey Petrakov, Alexander Panchenko, Timothy Baldwin, Preslav Nakov, Maxim Panov, Artem Shelmanov | cs.CL | cs.CL, cs.LG | Uncertainty quantification (UQ) is becoming increasingly recognized as a
critical component of applications that rely on machine learning (ML). The
rapid proliferation of large language models (LLMs) has stimulated researchers
to seek efficient and effective approaches to UQ in text generation tasks, as
in addition to ... | [
-0.1766001433,
-0.8027713895,
0.6182125807,
-0.0071425093,
-0.2064765841,
0.1904710084,
-0.260464251,
-0.3321168423,
-0.2883634269,
0.3381333351,
-0.2234049141,
-0.62748456,
-0.5667003393,
0.1136070192,
-0.4296094179,
0.6896702647,
-0.1009227335,
-0.0479934104,
-0.4123897851,
-... |
http://arxiv.org/abs/2406.15625v1 | http://arxiv.org/pdf/2406.15625v1 | Shortcomings of LLMs for Low-Resource Translation: Retrieval and Understanding are Both the Problem | 2024-06-21T20:02:22 | 2024-06-21T20:02:22 | Sara Court, Micha Elsner | cs.CL | cs.CL, cs.AI, cs.LG | This work investigates the in-context learning abilities of pretrained large
language models (LLMs) when instructed to translate text from a low-resource
language into a high-resource language as part of an automated machine
translation pipeline. We conduct a set of experiments translating Southern
Quechua to Spanish a... | [
-0.2787595391,
-0.6571136713,
0.7751449347,
0.0933216289,
-0.5635700226,
0.0179898664,
-0.5855321288,
-0.1858163029,
-0.1811237782,
0.5617429018,
-0.3946457803,
-0.3750956059,
-0.6831490397,
0.546492517,
-0.6649717093,
0.9374296069,
-0.1165437922,
0.0524868816,
-0.1606125385,
-... |
http://arxiv.org/abs/2407.00075v1 | http://arxiv.org/pdf/2407.00075v1 | Logicbreaks: A Framework for Understanding Subversion of Rule-based Inference | 2024-06-21T19:18:16 | 2024-06-21T19:18:16 | Anton Xue, Avishree Khare, Rajeev Alur, Surbhi Goel, Eric Wong | cs.AI | cs.AI, cs.CL, cs.CR, cs.LG | We study how to subvert language models from following the rules. We model
rule-following as inference in propositional Horn logic, a mathematical system
in which rules have the form "if $P$ and $Q$, then $R$" for some propositions
$P$, $Q$, and $R$. We prove that although transformers can faithfully abide by
such rule... | [
-0.2099972516,
-0.9273208976,
0.6600160599,
0.1401970536,
-0.0758269578,
-0.2876733243,
0.1377514601,
-0.2583591044,
0.1004048437,
1.1397724152,
-1.0148139,
-0.1106981412,
-0.5635180473,
-0.0721601844,
-0.7002076507,
1.1759266853,
0.1975605488,
-0.0957577005,
-0.101415202,
0.11... |
http://arxiv.org/abs/2406.15593v1 | http://arxiv.org/pdf/2406.15593v1 | News Deja Vu: Connecting Past and Present with Semantic Search | 2024-06-21T18:50:57 | 2024-06-21T18:50:57 | Brevin Franklin, Emily Silcock, Abhishek Arora, Tom Bryan, Melissa Dell | cs.CL | cs.CL, econ.GN, q-fin.EC | Social scientists and the general public often analyze contemporary events by
drawing parallels with the past, a process complicated by the vast, noisy, and
unstructured nature of historical texts. For example, hundreds of millions of
page scans from historical newspapers have been noisily transcribed.
Traditional spar... | [
-0.3184726536,
-0.9316216111,
0.9235084057,
-0.2357111275,
-0.4892499447,
0.208084181,
-0.0371636227,
-0.4218307137,
0.6111164689,
0.5541042089,
-0.6183183789,
-0.6661440134,
-0.6071597934,
-0.0053271018,
-0.5832258463,
0.6487786174,
-0.0594670065,
-0.0252671577,
-0.2270167619,
... |
http://arxiv.org/abs/2406.15586v1 | http://arxiv.org/pdf/2406.15586v1 | TinyStyler: Efficient Few-Shot Text Style Transfer with Authorship Embeddings | 2024-06-21T18:41:22 | 2024-06-21T18:41:22 | Zachary Horvitz, Ajay Patel, Kanishk Singh, Chris Callison-Burch, Kathleen McKeown, Zhou Yu | cs.CL | cs.CL | The goal of text style transfer is to transform the style of texts while
preserving their original meaning, often with only a few examples of the target
style. Existing style transfer methods generally rely on the few-shot
capabilities of large language models or on complex controllable text
generation approaches that ... | [
0.0693412349,
-0.6549715996,
0.7286179066,
0.0839579478,
-0.3243027627,
-0.017247729,
-0.4193862081,
-0.4661487937,
0.3248537481,
0.333401829,
-0.371628195,
-0.1458739489,
-0.6230230927,
0.2467672229,
-0.2059069425,
1.0922175646,
-0.0614214018,
-0.3032795787,
-0.1201637089,
-0.... |
http://arxiv.org/abs/2406.15583v1 | http://arxiv.org/pdf/2406.15583v1 | Detecting AI-Generated Text: Factors Influencing Detectability with Current Methods | 2024-06-21T18:31:49 | 2024-06-21T18:31:49 | Kathleen C. Fraser, Hillary Dawkins, Svetlana Kiritchenko | cs.CL | cs.CL, cs.CY | Large language models (LLMs) have advanced to a point that even humans have
difficulty discerning whether a text was generated by another human, or by a
computer. However, knowing whether a text was produced by human or artificial
intelligence (AI) is important to determining its trustworthiness, and has
applications i... | [
-0.3886434138,
-1.0849921703,
0.6537764668,
-0.0338435173,
-0.0982097164,
0.3717364967,
-0.0559490509,
-0.8526133895,
-0.1325752586,
0.4257138073,
-0.3323732316,
-0.4512526393,
-0.9999311566,
0.4339841306,
-0.9893988967,
0.8160534501,
-0.1627854258,
-0.3170884252,
-0.3541572094,
... |
http://arxiv.org/abs/2406.15576v1 | http://arxiv.org/pdf/2406.15576v1 | Contrastive Entity Coreference and Disambiguation for Historical Texts | 2024-06-21T18:22:14 | 2024-06-21T18:22:14 | Abhishek Arora, Emily Silcock, Leander Heldring, Melissa Dell | cs.CL | cs.CL, econ.GN, q-fin.EC | Massive-scale historical document collections are crucial for social science
research. Despite increasing digitization, these documents typically lack
unique cross-document identifiers for individuals mentioned within the texts,
as well as individual identifiers from external knowledgebases like
Wikipedia/Wikidata. Exi... | [
-0.6348504424,
-0.3987365365,
0.5234532356,
-0.373878032,
-0.3993911743,
0.2891186774,
-0.4889066815,
-0.5394485593,
0.151137203,
0.2055967003,
-0.1713915169,
-0.4469864666,
-0.6471155882,
0.3236695528,
-0.5093268752,
0.943338871,
-0.1049899682,
0.3982774615,
-0.4019203186,
-0.... |
http://arxiv.org/abs/2406.15570v1 | http://arxiv.org/pdf/2406.15570v1 | DEM: Distribution Edited Model for Training with Mixed Data Distributions | 2024-06-21T18:07:46 | 2024-06-21T18:07:46 | Dhananjay Ram, Aditya Rawal, Momchil Hardalov, Nikolaos Pappas, Sheng Zha | cs.CL | cs.CL, cs.LG, 68T50, F.2.2; I.2.7 | Training with mixed data distributions is a common and important part of
creating multi-task and instruction-following models. The diversity of the data
distributions and cost of joint training makes the optimization procedure
extremely challenging. Data mixing methods partially address this problem,
albeit having a su... | [
-0.7540103197,
-0.486518234,
0.2215533704,
-0.1115388647,
-0.6311966777,
0.0242906883,
0.0000148343,
-0.2984640598,
0.1308511943,
0.4960257113,
-0.887753129,
-0.5089362264,
-0.4698105156,
-0.0157905128,
-0.4071461558,
0.7138564587,
-0.0573928021,
0.4347954392,
-0.0359322429,
-0... |
http://arxiv.org/abs/2406.15567v1 | http://arxiv.org/pdf/2406.15567v1 | SAIL: Self-Improving Efficient Online Alignment of Large Language Models | 2024-06-21T18:05:35 | 2024-06-21T18:05:35 | Mucong Ding, Souradip Chakraborty, Vibhu Agrawal, Zora Che, Alec Koppel, Mengdi Wang, Amrit Bedi, Furong Huang | cs.LG | cs.LG, cs.AI, cs.CL, stat.ML | Reinforcement Learning from Human Feedback (RLHF) is a key method for
aligning large language models (LLMs) with human preferences. However, current
offline alignment approaches like DPO, IPO, and SLiC rely heavily on fixed
preference datasets, which can lead to sub-optimal performance. On the other
hand, recent litera... | [
-0.2175890803,
-0.2467948198,
0.3773267269,
0.2496542633,
-0.1895506531,
0.231090039,
-0.3654079139,
-0.2912795246,
0.3656214774,
0.5233174562,
-0.8485328555,
-0.2964368761,
-0.2344772518,
0.0172869638,
0.0895016342,
1.1779878139,
0.0019843597,
0.0940007716,
-0.1535336226,
-0.3... |
http://arxiv.org/abs/2406.15352v1 | http://arxiv.org/pdf/2406.15352v1 | A SMART Mnemonic Sounds like "Glue Tonic": Mixing LLMs with Student Feedback to Make Mnemonic Learning Stick | 2024-06-21T17:59:51 | 2024-06-21T17:59:51 | Nishant Balepur, Matthew Shu, Alexander Hoyle, Alison Robey, Shi Feng, Seraphina Goldfarb-Tarrant, Jordan Boyd-Graber | cs.CL | cs.CL | Keyword mnemonics are memorable explanations that link new terms to simpler
keywords. Prior works generate mnemonics for students, but they do not guide
models toward mnemonics students prefer and aid learning. We build SMART, a
mnemonic generator trained on feedback from real students learning new terms.
To train SMAR... | [
-0.1956968307,
-0.6332525015,
0.6504047513,
-0.1561608464,
-0.7454484105,
0.3220914602,
-0.1947307736,
-0.0459306762,
0.7002305984,
0.1624607891,
-0.5909559131,
-0.3936177492,
-0.2871746719,
-0.1445187181,
-0.0861132666,
1.5636457205,
0.3020988405,
0.2995461524,
-0.3571543396,
... |
http://arxiv.org/abs/2406.15334v1 | http://arxiv.org/pdf/2406.15334v1 | Multimodal Task Vectors Enable Many-Shot Multimodal In-Context Learning | 2024-06-21T17:50:02 | 2024-06-21T17:50:02 | Brandon Huang, Chancharik Mitra, Assaf Arbelle, Leonid Karlinsky, Trevor Darrell, Roei Herzig | cs.CV | cs.CV, cs.AI, cs.CL, cs.LG | The recent success of interleaved Large Multimodal Models (LMMs) in few-shot
learning suggests that in-context learning (ICL) with many examples can be
promising for learning new tasks. However, this many-shot multimodal ICL
setting has one crucial problem: it is fundamentally limited by the model's
context length set ... | [
-0.5372416377,
-0.5278092027,
0.4095258117,
0.2348082811,
-0.4509232342,
-0.204907164,
-0.3963748515,
-0.1756905913,
-0.0998937413,
-0.0375469141,
-0.8077046275,
-0.2975808084,
-0.717876792,
0.0291290674,
0.0169073045,
0.7643243074,
-0.0166880004,
0.0684156567,
-0.5485426188,
-... |
http://arxiv.org/abs/2406.15330v1 | http://arxiv.org/pdf/2406.15330v1 | Gradient-Mask Tuning Elevates the Upper Limits of LLM Performance | 2024-06-21T17:42:52 | 2024-06-21T17:42:52 | Haoling Li, Xin Zhang, Xiao Liu, Yeyun Gong, Yifan Wang, Yujiu Yang, Qi Chen, Peng Cheng | cs.AI | cs.AI, cs.CL | Large language models (LLMs) have revolutionized lots of fields of research.
Although it is well-known that fine-tuning is essential for enhancing the
capabilities of LLMs, existing research suggests that there is potential
redundancy in the fine-tuning process and therefore proposes to update only a
subset of paramete... | [
-0.6242321134,
-0.7252633572,
0.1935913712,
0.1570169479,
-0.3228862584,
0.3481296301,
-0.5237601399,
-0.5112364292,
0.0089321388,
0.420637995,
-0.5954406261,
-0.482465744,
-0.7654756308,
-0.234443292,
-0.1648261249,
1.0640727282,
0.0470387898,
-0.1555768996,
0.0579133183,
-0.1... |
http://arxiv.org/abs/2406.15319v2 | http://arxiv.org/pdf/2406.15319v2 | LongRAG: Enhancing Retrieval-Augmented Generation with Long-context LLMs | 2024-06-21T17:23:21 | 2024-06-30T15:01:36 | Ziyan Jiang, Xueguang Ma, Wenhu Chen | cs.CL | cs.CL, cs.AI | In traditional RAG framework, the basic retrieval units are normally short.
The common retrievers like DPR normally work with 100-word Wikipedia
paragraphs. Such a design forces the retriever to search over a large corpus to
find the `needle' unit. In contrast, the readers only need to extract answers
from the short re... | [
-0.7879315019,
-0.6246995926,
0.4418401718,
0.2038750052,
-0.5216950774,
0.0183820799,
-0.2925771773,
-0.6948705912,
0.7061244249,
0.5227595568,
-0.3528838754,
-0.2289272994,
-0.2876657844,
0.4513270259,
-0.5062505603,
1.4101748466,
-0.093756929,
-0.334446162,
-0.2340430468,
-0... |
http://arxiv.org/abs/2406.15313v1 | http://arxiv.org/pdf/2406.15313v1 | STARD: A Chinese Statute Retrieval Dataset with Real Queries Issued by Non-professionals | 2024-06-21T17:10:09 | 2024-06-21T17:10:09 | Weihang Su, Yiran Hu, Anzhe Xie, Qingyao Ai, Zibing Que, Ning Zheng, Yun Liu, Weixing Shen, Yiqun Liu | cs.IR | cs.IR, cs.CL | Statute retrieval aims to find relevant statutory articles for specific
queries. This process is the basis of a wide range of legal applications such
as legal advice, automated judicial decisions, legal document drafting, etc.
Existing statute retrieval benchmarks focus on formal and professional queries
from sources l... | [
-0.3780373037,
-0.5699380636,
0.391466558,
0.0515484773,
-0.6613647342,
-0.3037790656,
-0.2702268958,
-0.6314501762,
0.2597863972,
0.8901216388,
-0.179936707,
-0.5126386285,
-0.2598583996,
-0.2001102865,
-0.334027499,
0.8055361509,
0.3559083343,
0.2007439733,
0.1053936332,
-0.3... |
http://arxiv.org/abs/2406.15294v1 | http://arxiv.org/pdf/2406.15294v1 | NLP-KG: A System for Exploratory Search of Scientific Literature in Natural Language Processing | 2024-06-21T16:38:22 | 2024-06-21T16:38:22 | Tim Schopf, Florian Matthes | cs.CL | cs.CL | Scientific literature searches are often exploratory, whereby users are not
yet familiar with a particular field or concept but are interested in learning
more about it. However, existing systems for scientific literature search are
typically tailored to keyword-based lookup searches, limiting the possibilities
for exp... | [
-0.4970154762,
-0.5428873301,
0.4285010993,
0.2878428102,
-0.2101297826,
0.2680435479,
-0.0403166153,
-0.5453405976,
0.5572442412,
0.3903459907,
-0.4232903719,
-0.6938202381,
-0.647290349,
0.4244447947,
-0.4291984439,
0.976827085,
0.0978604108,
-0.1217103526,
-0.3595047891,
-0.... |
http://arxiv.org/abs/2406.15284v1 | http://arxiv.org/pdf/2406.15284v1 | The Greek podcast corpus: Competitive speech models for low-resourced languages with weakly supervised data | 2024-06-21T16:28:47 | 2024-06-21T16:28:47 | Georgios Paraskevopoulos, Chara Tsoukala, Athanasios Katsamanis, Vassilis Katsouros | cs.CL | cs.CL, cs.SD, eess.AS | The development of speech technologies for languages with limited digital
representation poses significant challenges, primarily due to the scarcity of
available data. This issue is exacerbated in the era of large, data-intensive
models. Recent research has underscored the potential of leveraging weak
supervision to au... | [
-0.3563038111,
-0.3645295501,
0.2679626942,
-0.1719912887,
-0.2536321878,
0.041343119,
-0.617056489,
-0.5843272805,
0.4223498106,
0.2330056727,
-0.340066433,
-0.4309372902,
-0.6860805154,
0.1615364403,
-0.5406709909,
0.8540324569,
0.3682383895,
0.3690167665,
-0.1057752892,
0.00... |
http://arxiv.org/abs/2406.15279v1 | http://arxiv.org/pdf/2406.15279v1 | Cross-Modality Safety Alignment | 2024-06-21T16:14:15 | 2024-06-21T16:14:15 | Siyin Wang, Xingsong Ye, Qinyuan Cheng, Junwen Duan, Shimin Li, Jinlan Fu, Xipeng Qiu, Xuanjing Huang | cs.AI | cs.AI, cs.CL | As Artificial General Intelligence (AGI) becomes increasingly integrated into
various facets of human life, ensuring the safety and ethical alignment of such
systems is paramount. Previous studies primarily focus on single-modality
threats, which may not suffice given the integrated and complex nature of
cross-modality... | [
0.0745874569,
-0.5803428292,
0.4756297767,
-0.206930846,
0.1146147922,
-0.0147194648,
0.3429086804,
-0.8154956698,
-0.4808572531,
0.1800240576,
-0.7032164335,
-0.1999257654,
-0.5544369817,
-0.0865732357,
-0.3296842575,
1.1246778965,
-0.0932467878,
0.1056651697,
-0.1903089136,
-... |
http://arxiv.org/abs/2406.15275v1 | http://arxiv.org/pdf/2406.15275v1 | Cognitive Map for Language Models: Optimal Planning via Verbally Representing the World Model | 2024-06-21T16:10:05 | 2024-06-21T16:10:05 | Doyoung Kim, Jongwon Lee, Jinho Park, Minjoon Seo | cs.CL | cs.CL | Language models have demonstrated impressive capabilities across various
natural language processing tasks, yet they struggle with planning tasks
requiring multi-step simulations. Inspired by human cognitive processes, this
paper investigates the optimal planning power of language models that can
construct a cognitive ... | [
-0.5579163432,
-0.811481297,
0.7215253115,
0.2556962371,
0.0856188983,
-0.149622798,
-0.6817561388,
-0.5485413074,
-0.0765767768,
0.8398501277,
-0.5861060619,
-0.5678155422,
-0.3373990059,
-0.4318810701,
-0.155507803,
0.669588387,
0.0706121102,
-0.2937891185,
-0.3193847239,
0.1... |
http://arxiv.org/abs/2406.15267v1 | http://arxiv.org/pdf/2406.15267v1 | Evaluating Diversity in Automatic Poetry Generation | 2024-06-21T16:03:21 | 2024-06-21T16:03:21 | Yanran Chen, Hannes Gröner, Sina Zarrieß, Steffen Eger | cs.CL | cs.CL | Natural Language Generation (NLG), and more generally generative AI, are
among the currently most impactful research fields. Creative NLG, such as
automatic poetry generation, is a fascinating niche in this area. While most
previous research has focused on forms of the Turing test when evaluating
automatic poetry gener... | [
-0.5616077185,
-0.5656726956,
0.2817209363,
0.6349990368,
-0.3485002518,
0.2389992326,
-0.3672321439,
-0.875025332,
0.0864090845,
0.2960561514,
-0.5774538517,
-0.5267344713,
-0.6792356968,
0.6451413631,
-0.6245039105,
1.0445498228,
0.1113808006,
-0.1797704399,
-0.1898656785,
-0... |
http://arxiv.org/abs/2406.15265v1 | http://arxiv.org/pdf/2406.15265v1 | Perception of Phonological Assimilation by Neural Speech Recognition Models | 2024-06-21T15:58:22 | 2024-06-21T15:58:22 | Charlotte Pouw, Marianne de Heer Kloots, Afra Alishahi, Willem Zuidema | cs.CL | cs.CL | Human listeners effortlessly compensate for phonological changes during
speech perception, often unconsciously inferring the intended sounds. For
example, listeners infer the underlying /n/ when hearing an utterance such as
"clea[m] pan", where [m] arises from place assimilation to the following labial
[p]. This articl... | [
-0.2523859143,
-0.308090657,
0.5280785561,
0.1481253952,
-0.1526816785,
-0.4506106675,
-0.2558533847,
-0.7697080374,
-0.1021850929,
0.4264301658,
-0.5353431702,
0.0300253797,
-0.5773752928,
-0.0213227998,
-0.36220029,
0.6127528548,
0.3783841133,
0.3281402588,
-0.0665077195,
-0.... |
http://arxiv.org/abs/2406.15264v1 | http://arxiv.org/pdf/2406.15264v1 | Towards Fine-Grained Citation Evaluation in Generated Text: A Comparative Analysis of Faithfulness Metrics | 2024-06-21T15:57:24 | 2024-06-21T15:57:24 | Weijia Zhang, Mohammad Aliannejadi, Yifei Yuan, Jiahuan Pei, Jia-Hong Huang, Evangelos Kanoulas | cs.IR | cs.IR, cs.CL | Large language models (LLMs) often produce unsupported or unverifiable
information, known as "hallucinations." To mitigate this, retrieval-augmented
LLMs incorporate citations, grounding the content in verifiable sources.
Despite such developments, manually assessing how well a citation supports the
associated statemen... | [
-0.3601643741,
-0.6080797315,
0.9805639386,
0.3330749571,
-0.2612007856,
0.1118416786,
-0.2334695905,
-0.5573391318,
0.221477896,
0.1761525273,
-0.0361297429,
-0.4180502892,
-0.5209408402,
0.4925999939,
-0.4748757482,
0.8608223796,
0.0066208979,
0.4352356195,
-0.5089601874,
-0.... |
http://arxiv.org/abs/2406.15245v1 | http://arxiv.org/pdf/2406.15245v1 | Unsupervised Morphological Tree Tokenizer | 2024-06-21T15:35:49 | 2024-06-21T15:35:49 | Qingyang Zhu, Xiang Hu, Pengyu Ji, Wei Wu, Kewei Tu | cs.CL | cs.CL, cs.LG | As a cornerstone in language modeling, tokenization involves segmenting text
inputs into pre-defined atomic units. Conventional statistical tokenizers often
disrupt constituent boundaries within words, thereby corrupting semantic
information. To address this drawback, we introduce morphological structure
guidance to to... | [
0.1185467765,
-0.7944399714,
0.3055815101,
0.0448058583,
-0.685565114,
0.210950613,
0.141249761,
-0.2728080153,
0.2576948404,
0.5762091875,
-0.3435899317,
-0.642346561,
-0.5095208883,
-0.0512203984,
-0.4328539968,
0.8336952329,
0.2736874521,
0.2357782871,
-0.0282315817,
0.31425... |
http://arxiv.org/abs/2406.15231v1 | http://arxiv.org/pdf/2406.15231v1 | Detecting Synthetic Lyrics with Few-Shot Inference | 2024-06-21T15:19:21 | 2024-06-21T15:19:21 | Yanis Labrak, Gabriel Meseguer-Brocal, Elena V. Epure | cs.CL | cs.CL, cs.AI, cs.LG | In recent years, generated content in music has gained significant
popularity, with large language models being effectively utilized to produce
human-like lyrics in various styles, themes, and linguistic structures. This
technological advancement supports artists in their creative processes but also
raises issues of au... | [
-0.5350911021,
-0.7121585608,
0.5604765415,
0.3735080063,
-0.2310885638,
0.1448646039,
-0.3035960793,
-0.4220003486,
0.2425367385,
0.8207525611,
-0.9439250231,
-0.8330945373,
-0.4903860092,
0.3681403697,
-0.5810675025,
0.9826133847,
0.2036350816,
-0.2450047582,
-0.2898766994,
-... |
http://arxiv.org/abs/2406.15227v1 | http://arxiv.org/pdf/2406.15227v1 | A LLM-Based Ranking Method for the Evaluation of Automatic Counter-Narrative Generation | 2024-06-21T15:11:33 | 2024-06-21T15:11:33 | Irune Zubiaga, Aitor Soroa, Rodrigo Agerri | cs.CL | cs.CL | The proliferation of misinformation and harmful narratives in online
discourse has underscored the critical need for effective Counter Narrative
(CN) generation techniques. However, existing automatic evaluation methods
often lack interpretability and fail to capture the nuanced relationship
between generated CNs and h... | [
-0.4299325347,
-0.8232606649,
0.4913452566,
-0.1153414547,
-0.3268568218,
0.3329496086,
-0.1660154164,
-0.5037764907,
0.3599822819,
0.4939459562,
-0.4422570169,
-0.6550405622,
-0.7072954774,
0.061400719,
-0.4526049197,
1.0760862827,
0.2374885082,
-0.098012507,
-0.1374850273,
-0... |
http://arxiv.org/abs/2406.15214v1 | http://arxiv.org/pdf/2406.15214v1 | Unsupervised Extraction of Dialogue Policies from Conversations | 2024-06-21T14:57:25 | 2024-06-21T14:57:25 | Makesh Narsimhan Sreedhar, Traian Rebedea, Christopher Parisien | cs.CL | cs.CL | Dialogue policies play a crucial role in developing task-oriented dialogue
systems, yet their development and maintenance are challenging and typically
require substantial effort from experts in dialogue modeling. While in many
situations, large amounts of conversational data are available for the task at
hand, people ... | [
-0.2944217026,
-0.6324490905,
0.5919755101,
0.1328449845,
-0.2375728488,
0.5216192603,
-0.3681583107,
-0.1227131784,
0.4054968655,
1.3555303812,
-0.748100996,
-0.4213885665,
-0.5908197165,
-0.1607468873,
-0.1929302216,
1.0562237501,
0.0068161841,
-0.0383425467,
-0.111029014,
0.... |
http://arxiv.org/abs/2406.15211v1 | http://arxiv.org/pdf/2406.15211v1 | How Effective is GPT-4 Turbo in Generating School-Level Questions from Textbooks Based on Bloom's Revised Taxonomy? | 2024-06-21T14:52:37 | 2024-06-21T14:52:37 | Subhankar Maity, Aniket Deroy, Sudeshna Sarkar | cs.CL | cs.CL, cs.AI | We evaluate the effectiveness of GPT-4 Turbo in generating educational
questions from NCERT textbooks in zero-shot mode. Our study highlights GPT-4
Turbo's ability to generate questions that require higher-order thinking
skills, especially at the "understanding" level according to Bloom's Revised
Taxonomy. While we fin... | [
-0.5474287868,
-0.9842044711,
0.6029254794,
-0.180652678,
-0.1278633177,
0.1641268581,
-0.235538885,
-0.3403467238,
-0.1663235128,
-0.08893843,
-0.4441893697,
-0.191077441,
-0.190920487,
0.3925519884,
-0.5048026443,
1.2620226145,
0.1969530582,
-0.2324101031,
-0.3908612728,
0.46... |
http://arxiv.org/abs/2406.15193v3 | http://arxiv.org/pdf/2406.15193v3 | Reward Steering with Evolutionary Heuristics for Decoding-time Alignment | 2024-06-21T14:35:16 | 2024-06-25T16:55:03 | Chia-Yu Hung, Navonil Majumder, Ambuj Mehrish, Soujanya Poria | cs.CL | cs.CL | The widespread applicability and increasing omnipresence of LLMs have
instigated a need to align LLM responses to user and stakeholder preferences.
Many preference optimization approaches have been proposed that fine-tune LLM
parameters to achieve good alignment. However, such parameter tuning is known
to interfere wit... | [
-0.1887152344,
-0.2386736572,
0.5606449842,
0.4096941054,
-0.4437006414,
0.4246044755,
-0.0912475064,
-0.5394085646,
0.4688648283,
0.2695671618,
-0.99487257,
-0.1465154141,
-0.6902165413,
0.0436229184,
0.0137996301,
0.9675782323,
-0.1875510961,
-0.108823292,
-0.0920570195,
-0.2... |
http://arxiv.org/abs/2406.15178v1 | http://arxiv.org/pdf/2406.15178v1 | Hybrid Alignment Training for Large Language Models | 2024-06-21T14:23:57 | 2024-06-21T14:23:57 | Chenglong Wang, Hang Zhou, Kaiyan Chang, Bei Li, Yongyu Mu, Tong Xiao, Tongran Liu, Jingbo Zhu | cs.CL | cs.CL | Alignment training is crucial for enabling large language models (LLMs) to
cater to human intentions and preferences. It is typically performed based on
two stages with different objectives: instruction-following alignment and
human-preference alignment. However, aligning LLMs with these objectives in
sequence suffers ... | [
-0.2870521247,
-0.4694188833,
0.4514427483,
0.1319782436,
-0.2793540359,
0.5612205863,
-0.5256299376,
-0.248535499,
0.4381592572,
0.3074340224,
-0.584373951,
-0.1681829244,
-0.3909149468,
0.0370506383,
-0.1258046925,
0.9498486519,
-0.5293749571,
0.2501702309,
-0.2282957435,
-0.... |
http://arxiv.org/abs/2406.15175v1 | http://arxiv.org/pdf/2406.15175v1 | Enhancing Idiomatic Representation in Multiple Languages via an Adaptive Contrastive Triplet Loss | 2024-06-21T14:21:41 | 2024-06-21T14:21:41 | Wei He, Marco Idiart, Carolina Scarton, Aline Villavicencio | cs.CL | cs.CL, cs.AI | Accurately modeling idiomatic or non-compositional language has been a
longstanding challenge in Natural Language Processing (NLP). This is partly
because these expressions do not derive their meanings solely from their
constituent words, but also due to the scarcity of relevant data resources, and
their impact on the ... | [
-0.2202448696,
-0.1915683299,
0.3318950534,
0.436786294,
-0.6526147723,
0.3384508491,
-0.2729121447,
-0.1839325279,
0.1617493778,
0.4160420597,
-0.4316839278,
-0.3500888646,
-0.6572122574,
0.1735781729,
-0.1155392453,
0.9263570905,
-0.5535632968,
0.0691495389,
-0.0473922156,
-0... |
http://arxiv.org/abs/2406.15534v1 | http://arxiv.org/pdf/2406.15534v1 | Geneverse: A collection of Open-source Multimodal Large Language Models for Genomic and Proteomic Research | 2024-06-21T14:19:10 | 2024-06-21T14:19:10 | Tianyu Liu, Yijia Xiao, Xiao Luo, Hua Xu, W. Jim Zheng, Hongyu Zhao | cs.LG | cs.LG, cs.AI, cs.CL, q-bio.QM | The applications of large language models (LLMs) are promising for biomedical
and healthcare research. Despite the availability of open-source LLMs trained
using a wide range of biomedical data, current research on the applications of
LLMs to genomics and proteomics is still limited. To fill this gap, we propose
a coll... | [
-0.3031856418,
-0.7068505287,
0.4436061382,
-0.0277273115,
0.0601135306,
0.0022061644,
-0.3864161074,
-0.169102028,
0.1465666294,
0.7452801466,
-0.6254812479,
-0.4712156951,
-0.6069226861,
0.4401339591,
-0.2790356278,
1.03469944,
0.0891692117,
0.609249115,
-0.4411175847,
-0.466... |
http://arxiv.org/abs/2406.15163v1 | http://arxiv.org/pdf/2406.15163v1 | A Syntax-Injected Approach for Faster and More Accurate Sentiment Analysis | 2024-06-21T14:08:25 | 2024-06-21T14:08:25 | Muhammad Imran, Olga Kellert, Carlos Gómez-Rodríguez | cs.CL | cs.CL | Sentiment Analysis (SA) is a crucial aspect of Natural Language Processing
(NLP), addressing subjective assessments in textual content. Syntactic parsing
is useful in SA because explicit syntactic information can improve accuracy
while providing explainability, but it tends to be a computational bottleneck
in practice ... | [
-0.4198669195,
-0.5170834661,
0.4609747827,
0.4256070256,
-0.4679597914,
0.2650905848,
-0.2377718687,
-0.3142981231,
0.3944448829,
0.4060953557,
-0.3050960004,
-0.5460765362,
-0.593436718,
0.3248617351,
-0.0919328406,
0.9740834236,
0.4376449287,
0.1418736875,
-0.1178992689,
-0.... |
http://arxiv.org/abs/2406.15130v1 | http://arxiv.org/pdf/2406.15130v1 | Assessing Good, Bad and Ugly Arguments Generated by ChatGPT: a New Dataset, its Methodology and Associated Tasks | 2024-06-21T13:27:10 | 2024-06-21T13:27:10 | Victor Hugo Nascimento Rocha, Igor Cataneo Silveira, Paulo Pirozelli, Denis Deratani Mauá, Fabio Gagliardi Cozman | cs.CL | cs.CL, cs.AI | The recent success of Large Language Models (LLMs) has sparked concerns about
their potential to spread misinformation. As a result, there is a pressing need
for tools to identify ``fake arguments'' generated by such models. To create
these tools, examples of texts generated by LLMs are needed. This paper
introduces a ... | [
-0.4383307695,
-1.136259675,
0.2470959723,
-0.1345236897,
-0.3790794909,
-0.1870717853,
-0.0355272815,
-0.1568965763,
-0.0960094258,
0.440023005,
-0.3560114801,
-0.3721679151,
-0.4764524102,
-0.0245701466,
-0.5440947413,
1.5097049475,
-0.1717681885,
0.3054790199,
-0.2909660637,
... |
http://arxiv.org/abs/2406.15111v2 | http://arxiv.org/pdf/2406.15111v2 | Investigating the impact of 2D gesture representation on co-speech gesture generation | 2024-06-21T12:59:20 | 2024-06-24T08:19:00 | Teo Guichoux, Laure Soulier, Nicolas Obin, Catherine Pelachaud | cs.AI | cs.AI, cs.CL, cs.CV | Co-speech gestures play a crucial role in the interactions between humans and
embodied conversational agents (ECA). Recent deep learning methods enable the
generation of realistic, natural co-speech gestures synchronized with speech,
but such approaches require large amounts of training data. "In-the-wild"
datasets, wh... | [
-0.3015093505,
-0.3651036918,
0.2173799276,
-0.1887181103,
-0.2805254161,
0.0065880041,
0.3143495619,
-0.8483166099,
0.4698776603,
0.0951043591,
-0.4171116352,
-0.5288847685,
-0.7564406991,
-0.0925321206,
-0.3631033897,
0.7770360112,
0.3047352433,
0.0222103391,
-0.3128623366,
-... |
http://arxiv.org/abs/2406.15109v1 | http://arxiv.org/pdf/2406.15109v1 | Brain-Like Language Processing via a Shallow Untrained Multihead Attention Network | 2024-06-21T12:54:03 | 2024-06-21T12:54:03 | Badr AlKhamissi, Greta Tuckute, Antoine Bosselut, Martin Schrimpf | cs.CL | cs.CL, cs.LG | Large Language Models (LLMs) have been shown to be effective models of the
human language system, with some models predicting most explainable variance of
brain activity in current datasets. Even in untrained models, the
representations induced by architectural priors can exhibit reasonable
alignment to brain data. In ... | [
-0.2188682407,
-0.5417765975,
0.4865683913,
0.1115693524,
-0.1003601924,
0.0441931076,
-0.4040586054,
-0.3866653144,
0.2087935358,
0.4466562569,
-0.4507463574,
-0.3813433647,
-0.2996991873,
-0.0699324757,
-0.2511769831,
0.779489994,
0.2555087507,
0.1233423203,
-0.0933342725,
-0... |
http://arxiv.org/abs/2406.15085v1 | http://arxiv.org/pdf/2406.15085v1 | A Unified Framework for Input Feature Attribution Analysis | 2024-06-21T12:01:03 | 2024-06-21T12:01:03 | Jingyi Sun, Pepa Atanasova, Isabelle Augenstein | cs.CL | cs.CL, 68T50, I.2.7 | Explaining the decision-making process of machine learning models is crucial
for ensuring their reliability and fairness. One popular explanation form
highlights key input features, such as i) tokens (e.g., Shapley Values and
Integrated Gradients), ii) interactions between tokens (e.g., Bivariate Shapley
and Attention-... | [
-0.425876379,
-0.6382103562,
0.8205219507,
0.2844071388,
-0.1253577173,
-0.2345041037,
-0.1378462613,
-0.3920050263,
0.0067422492,
0.4253829122,
-0.5311119556,
-0.2534650564,
-0.5159078836,
-0.2080173045,
-0.5597977638,
1.1539440155,
-0.4306121469,
0.0883048624,
-0.6010152102,
... |
http://arxiv.org/abs/2406.15066v1 | http://arxiv.org/pdf/2406.15066v1 | Cross-lingual paraphrase identification | 2024-06-21T11:37:24 | 2024-06-21T11:37:24 | Inessa Fedorova, Aleksei Musatow | cs.CL | cs.CL | The paraphrase identification task involves measuring semantic similarity
between two short sentences. It is a tricky task, and multilingual paraphrase
identification is even more challenging. In this work, we train a bi-encoder
model in a contrastive manner to detect hard paraphrases across multiple
languages. This ap... | [
-0.2065287977,
-0.6721403003,
0.4865733981,
0.3439258039,
-0.4645033479,
-0.102193892,
-0.0041279071,
-0.3090928197,
0.0553466342,
0.4713412821,
-0.2846126258,
-0.5155987144,
-0.725548923,
0.2987761199,
-0.4659532607,
0.7983224392,
-0.2639807463,
0.3657698631,
-0.3058062494,
-0... |
http://arxiv.org/abs/2406.15053v1 | http://arxiv.org/pdf/2406.15053v1 | PARIKSHA : A Large-Scale Investigation of Human-LLM Evaluator Agreement on Multilingual and Multi-Cultural Data | 2024-06-21T11:00:38 | 2024-06-21T11:00:38 | Ishaan Watts, Varun Gumma, Aditya Yadavalli, Vivek Seshadri, Manohar Swaminathan, Sunayana Sitaram | cs.CL | cs.CL | Evaluation of multilingual Large Language Models (LLMs) is challenging due to
a variety of factors -- the lack of benchmarks with sufficient linguistic
diversity, contamination of popular benchmarks into LLM pre-training data and
the lack of local, cultural nuances in translated benchmarks. In this work, we
study human... | [
-0.4798421264,
-0.5805473328,
0.3673893809,
0.6057078838,
-0.3786070347,
0.349181056,
-0.6216073036,
-0.4709156752,
-0.0112895984,
0.4887569547,
-0.0548840538,
-0.4454658628,
-0.7155652046,
0.529145956,
-0.131800428,
0.6256456971,
-0.2861116827,
0.5484326482,
-0.2221192122,
-0.... |
http://arxiv.org/abs/2406.15050v1 | http://arxiv.org/pdf/2406.15050v1 | Tri-VQA: Triangular Reasoning Medical Visual Question Answering for Multi-Attribute Analysis | 2024-06-21T10:50:55 | 2024-06-21T10:50:55 | Lin Fan, Xun Gong, Cenyang Zheng, Yafei Ou | cs.LG | cs.LG, cs.AI, cs.CL, cs.CV, I.2.7; I.2.10; J.3 | The intersection of medical Visual Question Answering (Med-VQA) is a
challenging research topic with advantages including patient engagement and
clinical expert involvement for second opinions. However, existing Med-VQA
methods based on joint embedding fail to explain whether their provided results
are based on correct... | [
-0.6953495741,
-0.8608080745,
0.7910032272,
-0.0680542737,
-0.3623993099,
0.0881027803,
0.3131301701,
-0.4979406297,
0.3135905564,
0.4867405295,
-0.5183654428,
-0.3901422322,
-0.3949078918,
0.2507469058,
-0.2898474932,
0.653753221,
-0.1595682502,
0.5453813076,
-0.4050833881,
-0... |
http://arxiv.org/abs/2406.15045v1 | http://arxiv.org/pdf/2406.15045v1 | Harnessing Knowledge Retrieval with Large Language Models for Clinical Report Error Correction | 2024-06-21T10:48:21 | 2024-06-21T10:48:21 | Jinge Wu, Zhaolong Wu, Abul Hasan, Yunsoo Kim, Jason P. Y. Cheung, Teng Zhang, Honghan Wu | cs.CL | cs.CL | This study proposes an approach for error correction in clinical radiology
reports, leveraging large language models (LLMs) and retrieval-augmented
generation (RAG) techniques. The proposed framework employs internal and
external retrieval mechanisms to extract relevant medical entities and
relations from the report an... | [
0.4091005325,
-0.5411657095,
0.9664633274,
-0.0012607912,
-0.0047299163,
0.1103720441,
-0.2914273441,
-0.6055324674,
0.0655712113,
0.6067253351,
0.0077800392,
-0.4321447313,
-0.8072057962,
0.1450480968,
-0.1682514697,
0.6595715284,
-0.3310738802,
0.6882601976,
-0.1780382544,
-0... |
http://arxiv.org/abs/2406.15038v1 | http://arxiv.org/pdf/2406.15038v1 | Online detection and infographic explanation of spam reviews with data drift adaptation | 2024-06-21T10:35:46 | 2024-06-21T10:35:46 | Francisco de Arriba-Pérez, Silvia García-Méndez, Fátima Leal, Benedita Malheiro, J. C. Burguillo | cs.LG | cs.LG, cs.AI, cs.CL, cs.SI | Spam reviews are a pervasive problem on online platforms due to its
significant impact on reputation. However, research into spam detection in data
streams is scarce. Another concern lies in their need for transparency.
Consequently, this paper addresses those problems by proposing an online
solution for identifying an... | [
-0.8425768018,
-0.7982461452,
0.4162274003,
0.419521898,
-0.4797866344,
0.2552100122,
0.1428873241,
-0.361240834,
0.3455799222,
0.6854968071,
-1.0043094158,
-0.5457540751,
-0.5368191004,
-0.0222150367,
-0.5536623001,
1.6360191107,
-0.1098534316,
0.0173989478,
0.0031561675,
-0.1... |
http://arxiv.org/abs/2406.15032v1 | http://arxiv.org/pdf/2406.15032v1 | GiusBERTo: A Legal Language Model for Personal Data De-identification in Italian Court of Auditors Decisions | 2024-06-21T10:25:26 | 2024-06-21T10:25:26 | Giulio Salierno, Rosamaria Bertè, Luca Attias, Carla Morrone, Dario Pettazzoni, Daniela Battisti | cs.CL | cs.CL, cs.AI | Recent advances in Natural Language Processing have demonstrated the
effectiveness of pretrained language models like BERT for a variety of
downstream tasks. We present GiusBERTo, the first BERT-based model specialized
for anonymizing personal data in Italian legal documents. GiusBERTo is trained
on a large dataset of ... | [
-0.2438598722,
-0.8137440085,
0.5138777494,
-0.0677828193,
-0.3553614616,
-0.053630583,
-0.337872684,
-0.7577616572,
0.2411613762,
0.3757858872,
-0.4430032372,
-0.6021467447,
-0.6081548929,
-0.0518902354,
-0.910443604,
1.0227937698,
0.3922428489,
0.4983756542,
0.3169621825,
-0.... |
http://arxiv.org/abs/2406.15019v1 | http://arxiv.org/pdf/2406.15019v1 | MedOdyssey: A Medical Domain Benchmark for Long Context Evaluation Up to 200K Tokens | 2024-06-21T09:46:57 | 2024-06-21T09:46:57 | Yongqi Fan, Hongli Sun, Kui Xue, Xiaofan Zhang, Shaoting Zhang, Tong Ruan | cs.CL | cs.CL | Numerous advanced Large Language Models (LLMs) now support context lengths up
to 128K, and some extend to 200K. Some benchmarks in the generic domain have
also followed up on evaluating long-context capabilities. In the medical
domain, tasks are distinctive due to the unique contexts and need for domain
expertise, nece... | [
-0.1331279129,
-0.492941767,
0.9686886072,
0.1438353956,
-0.4156438112,
-0.1132669225,
-0.4288845658,
-0.6758257151,
0.2913212776,
0.444970578,
-0.6972950101,
-0.5805943608,
-0.5289389491,
0.3033874333,
-0.0992758498,
1.0996162891,
-0.2567447722,
0.1555345505,
-0.4052016437,
-0... |
http://arxiv.org/abs/2406.15015v1 | http://arxiv.org/pdf/2406.15015v1 | GraLMatch: Matching Groups of Entities with Graphs and Language Models | 2024-06-21T09:44:16 | 2024-06-21T09:44:16 | Fernando De Meer Pardo, Claude Lehmann, Dennis Gehrig, Andrea Nagy, Stefano Nicoli, Branka Hadji Misheva, Martin Braschler, Kurt Stockinger | cs.DB | cs.DB, cs.AI, cs.CL | In this paper, we present an end-to-end multi-source Entity Matching problem,
which we call entity group matching, where the goal is to assign to the same
group, records originating from multiple data sources but representing the same
real-world entity. We focus on the effects of transitively matched records,
i.e. the ... | [
-0.2279525399,
-0.7216991186,
0.4957809746,
-0.0121725658,
-0.0781781003,
0.4608208835,
-0.0020970185,
-0.5903376341,
0.1817084998,
0.6346296072,
-0.4837689698,
-0.4260567129,
-0.619700551,
-0.2334529161,
-0.4619655013,
1.0576697588,
-0.1413754076,
-0.1116272137,
-0.0501429141,
... |
http://arxiv.org/abs/2406.15000v1 | http://arxiv.org/pdf/2406.15000v1 | Unveiling the Impact of Multi-Modal Interactions on User Engagement: A Comprehensive Evaluation in AI-driven Conversations | 2024-06-21T09:26:55 | 2024-06-21T09:26:55 | Lichao Zhang, Jia Yu, Shuai Zhang, Long Li, Yangyang Zhong, Guanbao Liang, Yuming Yan, Qing Ma, Fangsheng Weng, Fayu Pan, Jing Li, Renjun Xu, Zhenzhong Lan | cs.CL | cs.CL, cs.AI | Large Language Models (LLMs) have significantly advanced user-bot
interactions, enabling more complex and coherent dialogues. However, the
prevalent text-only modality might not fully exploit the potential for
effective user engagement. This paper explores the impact of multi-modal
interactions, which incorporate image... | [
-0.4251136184,
-0.9871293306,
0.6304603815,
-0.0154843982,
-0.1974808425,
0.1897645444,
-0.3659606576,
-0.7653254867,
0.1118246093,
0.0934721753,
-0.4925145805,
-0.0593459457,
-0.3801992834,
-0.0560720228,
-0.1383837163,
0.5520972013,
0.3163438439,
0.0374146178,
-0.5491618514,
... |
http://arxiv.org/abs/2406.14993v1 | http://arxiv.org/pdf/2406.14993v1 | Disability Representations: Finding Biases in Automatic Image Generation | 2024-06-21T09:12:31 | 2024-06-21T09:12:31 | Yannis Tevissen | cs.CV | cs.CV, cs.CL | Recent advancements in image generation technology have enabled widespread
access to AI-generated imagery, prominently used in advertising, entertainment,
and progressively in every form of visual content. However, these technologies
often perpetuate societal biases. This study investigates the representation
biases in... | [
-0.7047746778,
-0.1851469129,
0.4505171478,
0.1934954077,
-0.2186274976,
-0.2788295746,
0.3157889545,
-0.7218242288,
0.4237688184,
0.3194555938,
-0.4736782908,
-0.0924339294,
-0.9235864282,
0.3387548327,
-0.1505652815,
0.5594137311,
-0.0968985334,
0.3037186861,
-0.2408788204,
-... |
http://arxiv.org/abs/2406.14991v1 | http://arxiv.org/pdf/2406.14991v1 | SpreadsheetBench: Towards Challenging Real World Spreadsheet Manipulation | 2024-06-21T09:06:45 | 2024-06-21T09:06:45 | Zeyao Ma, Bohan Zhang, Jing Zhang, Jifan Yu, Xiaokang Zhang, Xiaohan Zhang, Sijia Luo, Xi Wang, Jie Tang | cs.CL | cs.CL, cs.SE | We introduce SpreadsheetBench, a challenging spreadsheet manipulation
benchmark exclusively derived from real-world scenarios, designed to immerse
current large language models (LLMs) in the actual workflow of spreadsheet
users. Unlike existing benchmarks that rely on synthesized queries and
simplified spreadsheet file... | [
-0.1608174145,
-1.1620651484,
0.4240086079,
0.2587141395,
0.2060006112,
0.415699929,
-0.466321975,
-0.7006081343,
-0.1764806509,
0.6226522326,
-0.8485611081,
-0.5861503482,
-0.3820859194,
0.1137741357,
-0.6346581578,
0.7861561179,
-0.3607687354,
0.1355772465,
-0.4132698774,
-0.... |
http://arxiv.org/abs/2406.14986v2 | http://arxiv.org/pdf/2406.14986v2 | Do Large Language Models Exhibit Cognitive Dissonance? Studying the Difference Between Revealed Beliefs and Stated Answers | 2024-06-21T08:56:35 | 2024-07-02T14:02:04 | Manuel Mondal, Ljiljana Dolamic, Gérôme Bovet, Philippe Cudré-Mauroux, Julien Audiffren | cs.AI | cs.AI, cs.CL | Prompting and Multiple Choices Questions (MCQ) have become the preferred
approach to assess the capabilities of Large Language Models (LLMs), due to
their ease of manipulation and evaluation. Such experimental appraisals have
pointed toward the LLMs' apparent ability to perform causal reasoning or to
grasp uncertainty.... | [
-0.4635779858,
-0.7954386473,
1.2650977373,
0.2123142332,
-0.1907440275,
0.1472804844,
-0.3499258757,
-0.1260274649,
-0.2114658654,
0.3178747296,
-0.6687540412,
-0.2417882681,
-0.9436668754,
0.4614054561,
-0.3497949839,
0.6133282185,
-0.1137995943,
-0.0503412187,
-0.4850944579,
... |
http://arxiv.org/abs/2407.00072v2 | http://arxiv.org/pdf/2407.00072v2 | Pistis-RAG: A Scalable Cascading Framework Towards Content-Centric Retrieval-Augmented Generation | 2024-06-21T08:52:11 | 2024-07-03T06:54:16 | Yu Bai, Yukai Miao, Li Chen, Dan Li, Yanyu Ren, Hongtao Xie, Ce Yang, Xuhui Cai | cs.IR | cs.IR, cs.CL | In Greek mythology, Pistis symbolized good faith, trust, and reliability.
Drawing inspiration from these principles, Pistis-RAG is a scalable multi-stage
framework designed to address the challenges of large-scale retrieval-augmented
generation (RAG) systems. This framework consists of distinct stages: matching,
pre-ra... | [
-0.7592122555,
-0.39807567,
0.6024116278,
0.0621670373,
-0.2907767594,
0.2510458231,
-0.3135712147,
-0.1886115074,
0.4923233688,
0.3297588825,
-0.383197993,
-0.2508618236,
-0.7808841467,
-0.0435503498,
-0.479110986,
1.2122942209,
0.3395375907,
-0.1032256112,
-0.0410190597,
-0.0... |
http://arxiv.org/abs/2406.14979v1 | http://arxiv.org/pdf/2406.14979v1 | Retrieve-Plan-Generation: An Iterative Planning and Answering Framework for Knowledge-Intensive LLM Generation | 2024-06-21T08:45:52 | 2024-06-21T08:45:52 | Yuanjie Lyu, Zihan Niu, Zheyong Xie, Chao Zhang, Tong Xu, Yang Wang, Enhong Chen | cs.CL | cs.CL | Despite the significant progress of large language models (LLMs) in various
tasks, they often produce factual errors due to their limited internal
knowledge. Retrieval-Augmented Generation (RAG), which enhances LLMs with
external knowledge sources, offers a promising solution. However, these methods
can be misled by ir... | [
-0.2660461962,
-0.6422587633,
0.785815835,
0.3966379166,
-0.0522086285,
0.0781508982,
-0.1439784467,
-0.071281679,
0.1079045683,
0.8419541121,
-0.9415875077,
-0.3538566232,
-0.354074657,
-0.0468769111,
-0.1934546381,
0.7259161472,
-0.2181776315,
-0.1545667052,
-0.2870982289,
0.... |
http://arxiv.org/abs/2406.14972v1 | http://arxiv.org/pdf/2406.14972v1 | A Tale of Trust and Accuracy: Base vs. Instruct LLMs in RAG Systems | 2024-06-21T08:31:02 | 2024-06-21T08:31:02 | Florin Cuconasu, Giovanni Trappolini, Nicola Tonellotto, Fabrizio Silvestri | cs.CL | cs.CL, cs.IR | Retrieval Augmented Generation (RAG) represents a significant advancement in
artificial intelligence combining a retrieval phase with a generative phase,
with the latter typically being powered by large language models (LLMs). The
current common practices in RAG involve using "instructed" LLMs, which are
fine-tuned wit... | [
-0.2274751663,
-0.50390625,
0.7646620274,
-0.0874050111,
-0.1615576148,
0.6466357112,
-0.0276892222,
-0.5773251653,
0.0202412009,
0.7114585638,
-0.1702701896,
-0.2435908616,
-0.9566809535,
0.0223213322,
-0.4746319354,
1.0958566666,
0.0146038076,
0.216033712,
-0.4902861416,
-0.1... |
http://arxiv.org/abs/2406.14971v1 | http://arxiv.org/pdf/2406.14971v1 | Domain Adaptation of Llama3-70B-Instruct through Continual Pre-Training and Model Merging: A Comprehensive Evaluation | 2024-06-21T08:29:31 | 2024-06-21T08:29:31 | Shamane Siriwardhana, Mark McQuade, Thomas Gauthier, Lucas Atkins, Fernando Fernandes Neto, Luke Meyers, Anneketh Vij, Tyler Odenthal, Charles Goddard, Mary MacCarthy, Jacob Solawetz | cs.CL | cs.CL, cs.AI, cs.LG | We conducted extensive experiments on domain adaptation of the
Meta-Llama-3-70B-Instruct model on SEC data, exploring its performance on both
general and domain-specific benchmarks. Our focus included continual
pre-training (CPT) and model merging, aiming to enhance the model's
domain-specific capabilities while mitiga... | [
-0.2106537819,
-0.7164517641,
0.330218792,
0.3573635817,
-0.4786489606,
0.2951564789,
0.0028688733,
-0.8275248408,
-0.0699479729,
0.7342967987,
-0.3213590384,
-0.3609580398,
-0.6105720401,
-0.0269892924,
-0.547986269,
1.257469058,
-0.3433742523,
0.1317991763,
-0.1282211393,
0.2... |
http://arxiv.org/abs/2406.14956v1 | http://arxiv.org/pdf/2406.14956v1 | Unlocking the Global Synergies in Low-Rank Adapters | 2024-06-21T08:10:03 | 2024-06-21T08:10:03 | Zixi Zhang, Cheng Zhang, Xitong Gao, Robert D. Mullins, George A. Constantinides, Yiren Zhao | cs.LG | cs.LG, cs.CL | Low-rank Adaption (LoRA) has been the de-facto parameter-efficient
fine-tuning technique for large language models. We present HeteroLoRA, a
light-weight search algorithm that leverages zero-cost proxies to allocate the
limited LoRA trainable parameters across the model for better fine-tuned
performance. In addition to... | [
-0.2916663885,
-0.5279178619,
0.5896515846,
0.0593466684,
-0.2630868852,
-0.3185519576,
-0.2726217508,
-0.5837507248,
0.4172480404,
0.2066080421,
-0.5216752291,
-0.2646652162,
-0.3266915977,
-0.1307713091,
0.20886226,
0.7635840178,
0.2132245004,
-0.1081338301,
0.3756498694,
-0.... |
http://arxiv.org/abs/2406.14955v1 | http://arxiv.org/pdf/2406.14955v1 | ICLEval: Evaluating In-Context Learning Ability of Large Language Models | 2024-06-21T08:06:10 | 2024-06-21T08:06:10 | Wentong Chen, Yankai Lin, ZhenHao Zhou, HongYun Huang, Yantao Jia, Zhao Cao, Ji-Rong Wen | cs.CL | cs.CL | In-Context Learning (ICL) is a critical capability of Large Language Models
(LLMs) as it empowers them to comprehend and reason across interconnected
inputs. Evaluating the ICL ability of LLMs can enhance their utilization and
deepen our understanding of how this ability is acquired at the training stage.
However, exis... | [
-0.2173192501,
-0.4377467632,
0.4722416997,
0.415443182,
-0.3443759382,
0.3295472562,
-0.2280909568,
-0.4411367774,
-0.6853616238,
0.1336157471,
-0.1789128631,
-0.6004828215,
-0.6026610136,
-0.0411272235,
-0.7136697173,
1.1487531662,
-0.1353786588,
0.2801099718,
-0.6475662589,
... |
http://arxiv.org/abs/2406.14952v2 | http://arxiv.org/pdf/2406.14952v2 | ESC-Eval: Evaluating Emotion Support Conversations in Large Language Models | 2024-06-21T08:03:33 | 2024-06-24T12:24:52 | Haiquan Zhao, Lingyu Li, Shisong Chen, Shuqi Kong, Jiaan Wang, Kexin Huang, Tianle Gu, Yixu Wang, Dandan Liang, Zhixu Li, Yan Teng, Yanghua Xiao, Yingchun Wang | cs.CL | cs.CL | Emotion Support Conversation (ESC) is a crucial application, which aims to
reduce human stress, offer emotional guidance, and ultimately enhance human
mental and physical well-being. With the advancement of Large Language Models
(LLMs), many researchers have employed LLMs as the ESC models. However, the
evaluation of t... | [
-0.4879598618,
-0.8587347865,
0.7508326769,
0.1274136305,
-0.0950605199,
0.185654372,
-0.3572893739,
-0.5141678452,
0.2604905069,
0.3947345912,
-0.6018318534,
-0.7901230454,
-0.5441224575,
0.2834693193,
-0.4707544446,
0.9975271821,
0.0178590249,
0.1203559488,
-0.5595921278,
-0.... |
http://arxiv.org/abs/2406.14938v1 | http://arxiv.org/pdf/2406.14938v1 | Towards Retrieval Augmented Generation over Large Video Libraries | 2024-06-21T07:52:01 | 2024-06-21T07:52:01 | Yannis Tevissen, Khalil Guetari, Frédéric Petitpont | cs.CL | cs.CL, cs.AI | Video content creators need efficient tools to repurpose content, a task that
often requires complex manual or automated searches. Crafting a new video from
large video libraries remains a challenge. In this paper we introduce the task
of Video Library Question Answering (VLQA) through an interoperable
architecture tha... | [
-0.6482527852,
-0.7939049602,
0.4610512555,
-0.0232566278,
-0.1081306636,
0.3194471896,
0.1794532239,
-0.2115233243,
0.3183612823,
0.8606543541,
-0.6406528354,
0.0006634175,
-0.3039012849,
-0.1192166954,
-0.8420901895,
0.6950123906,
-0.0252952091,
-0.0401487052,
-0.3423191011,
... |
http://arxiv.org/abs/2406.15527v1 | http://arxiv.org/pdf/2406.15527v1 | Data Efficient Evaluation of Large Language Models and Text-to-Image Models via Adaptive Sampling | 2024-06-21T07:38:55 | 2024-06-21T07:38:55 | Cong Xu, Gayathri Saranathan, Mahammad Parwez Alam, Arpit Shah, James Lim, Soon Yee Wong, Foltin Martin, Suparna Bhattacharya | cs.LG | cs.LG, cs.CL | Evaluating LLMs and text-to-image models is a computationally intensive task
often overlooked. Efficient evaluation is crucial for understanding the diverse
capabilities of these models and enabling comparisons across a growing number
of new models and benchmarks. To address this, we introduce SubLIME, a
data-efficient... | [
-0.0440734737,
-0.7018559575,
0.6271313429,
0.4147569835,
-0.196799472,
0.3888244927,
-0.5810775757,
-0.0163459498,
-0.1219023094,
0.3346353769,
-0.2439791262,
-0.9306522608,
-0.5274295807,
0.0967535451,
-0.712382555,
0.5902468562,
0.1075671539,
0.0299417712,
-0.4078620076,
-0.... |
http://arxiv.org/abs/2406.14928v1 | http://arxiv.org/pdf/2406.14928v1 | Autonomous Agents for Collaborative Task under Information Asymmetry | 2024-06-21T07:37:19 | 2024-06-21T07:37:19 | Wei Liu, Chenxi Wang, Yifei Wang, Zihao Xie, Rennai Qiu, Yufan Dang, Zhuoyun Du, Weize Chen, Cheng Yang, Chen Qian | cs.AI | cs.AI, cs.CL, cs.HC, cs.MA, cs.SI | Large Language Model Multi-Agent Systems (LLM-MAS) have achieved great
progress in solving complex tasks. It performs communication among agents
within the system to collaboratively solve tasks, under the premise of shared
information. However, when agents' communication is leveraged to enhance human
cooperation, a new... | [
-0.4961181879,
-0.4323234558,
0.5713716149,
0.4276954532,
-0.1226061359,
0.2398780882,
-0.5318354368,
-0.6358495355,
0.682487607,
0.3676010072,
-0.8608220816,
-0.3032365441,
-0.6291683316,
0.0160397235,
-0.3796986043,
0.6612926126,
-0.0102106892,
0.0916124731,
-0.4107967317,
-0... |
http://arxiv.org/abs/2406.14917v1 | http://arxiv.org/pdf/2406.14917v1 | LLM2FEA: Discover Novel Designs with Generative Evolutionary Multitasking | 2024-06-21T07:20:51 | 2024-06-21T07:20:51 | Melvin Wong, Jiao Liu, Thiago Rios, Stefan Menzel, Yew Soon Ong | cs.AI | cs.AI, cs.CL, cs.CV, cs.LG, cs.NE | The rapid research and development of generative artificial intelligence has
enabled the generation of high-quality images, text, and 3D models from text
prompts. This advancement impels an inquiry into whether these models can be
leveraged to create digital artifacts for both creative and engineering
applications. Dra... | [
-0.571622014,
-0.7836910486,
0.7882028818,
0.5107762814,
-0.1316704154,
0.3634328544,
0.4772124588,
-0.7054317594,
0.3719320297,
0.058187183,
-0.9599505067,
-0.2263439596,
-0.4829242527,
0.0872642547,
-0.2731508315,
0.3448733985,
-0.263514936,
-0.2047040015,
-0.3814492524,
0.11... |
http://arxiv.org/abs/2406.14909v1 | http://arxiv.org/pdf/2406.14909v1 | MoA: Mixture of Sparse Attention for Automatic Large Language Model Compression | 2024-06-21T06:58:37 | 2024-06-21T06:58:37 | Tianyu Fu, Haofeng Huang, Xuefei Ning, Genghan Zhang, Boju Chen, Tianqi Wu, Hongyi Wang, Zixiao Huang, Shiyao Li, Shengen Yan, Guohao Dai, Huazhong Yang, Yu Wang | cs.LG | cs.LG, cs.AI, cs.CL, I.2.7 | Sparse attention can effectively mitigate the significant memory and
throughput demands of Large Language Models (LLMs) in long contexts. Existing
methods typically employ a uniform sparse attention mask, applying the same
sparse pattern across different attention heads and input lengths. However,
this uniform approach... | [
-0.5832493305,
-0.9452795386,
0.7250292897,
0.5573064089,
-0.4353843629,
-0.0543490537,
-0.4283888042,
-0.3945888877,
0.5656952262,
0.5684674382,
-0.5813429952,
-0.1519115269,
-0.6972932816,
0.0239437837,
-0.3273178637,
0.5083463788,
0.1855757982,
-0.3160063326,
-0.2736290395,
... |
http://arxiv.org/abs/2406.14898v2 | http://arxiv.org/pdf/2406.14898v2 | Safely Learning with Private Data: A Federated Learning Framework for Large Language Model | 2024-06-21T06:43:15 | 2024-06-26T04:28:38 | JiaYing Zheng, HaiNan Zhang, LingXiang Wang, WangJie Qiu, HongWei Zheng, ZhiMing Zheng | cs.CR | cs.CR, cs.CL | Private data, being larger and quality-higher than public data, can greatly
improve large language models (LLM). However, due to privacy concerns, this
data is often dispersed in multiple silos, making its secure utilization for
LLM training a challenge. Federated learning (FL) is an ideal solution for
training models ... | [
-0.1701172888,
-0.7549245358,
0.201310724,
0.317230314,
-0.0590665415,
0.3371030092,
-0.2933695018,
-0.3403322697,
-0.2534388602,
0.4877077937,
-0.4077344239,
-0.6454769969,
-0.4627702236,
-0.0726565346,
-0.3527254164,
1.0401474237,
0.095337525,
0.2804726362,
-0.3240596354,
-0.... |
http://arxiv.org/abs/2406.14894v1 | http://arxiv.org/pdf/2406.14894v1 | Talking the Talk Does Not Entail Walking the Walk: On the Limits of Large Language Models in Lexical Entailment Recognition | 2024-06-21T06:30:16 | 2024-06-21T06:30:16 | Candida M. Greco, Lucio La Cava, Andrea Tagarelli | cs.CL | cs.CL, cs.AI, cs.CY, cs.IR, physics.soc-ph | Verbs form the backbone of language, providing the structure and meaning to
sentences. Yet, their intricate semantic nuances pose a longstanding challenge.
Understanding verb relations through the concept of lexical entailment is
crucial for comprehending sentence meanings and grasping verb dynamics. This
work investig... | [
-0.3919551671,
-0.847494483,
1.0670266151,
-0.145587191,
-0.2984418571,
-0.3096676171,
-0.2881976664,
-0.5265315175,
0.2349379659,
0.517889142,
-0.6266447306,
-0.4226218462,
-0.84477669,
0.2002460957,
-0.5325629115,
0.5014638901,
0.0811230242,
-0.0456782132,
-0.2633481324,
-0.1... |
http://arxiv.org/abs/2406.14891v1 | http://arxiv.org/pdf/2406.14891v1 | Generate-then-Ground in Retrieval-Augmented Generation for Multi-hop Question Answering | 2024-06-21T06:26:38 | 2024-06-21T06:26:38 | Zhengliang Shi, Shuo Zhang, Weiwei Sun, Shen Gao, Pengjie Ren, Zhumin Chen, Zhaochun Ren | cs.CL | cs.CL, cs.IR | Multi-Hop Question Answering (MHQA) tasks present a significant challenge for
large language models (LLMs) due to the intensive knowledge required. Current
solutions, like Retrieval-Augmented Generation, typically retrieve potential
documents from an external corpus to read an answer. However, the performance
of this r... | [
-0.377525717,
-0.9446353316,
0.7499592304,
-0.0245760027,
-0.2135224342,
0.3994786739,
-0.0143045252,
-0.177407071,
0.1378138661,
0.3906457126,
-0.838365674,
-0.1317738742,
-0.6012247205,
0.0878797024,
-0.3384924829,
0.8579018116,
0.0987963378,
-0.1443197578,
-0.3419657648,
0.0... |
http://arxiv.org/abs/2406.14890v1 | http://arxiv.org/pdf/2406.14890v1 | InterBiasing: Boost Unseen Word Recognition through Biasing Intermediate Predictions | 2024-06-21T06:25:10 | 2024-06-21T06:25:10 | Yu Nakagome, Michael Hentschel | cs.CL | cs.CL, eess.AS | Despite recent advances in end-to-end speech recognition methods, their
output is biased to the training data's vocabulary, resulting in inaccurate
recognition of unknown terms or proper nouns. To improve the recognition
accuracy for a given set of such terms, we propose an adaptation parameter-free
approach based on S... | [
-0.4523718655,
-0.4486984313,
0.3935499787,
0.3292172551,
-0.5503059626,
0.2168425769,
-0.4368026555,
-0.3234520853,
0.2929298878,
0.3150542974,
-0.619509697,
-0.4017384052,
-0.74515903,
0.1417422146,
-0.3867739141,
1.0657244921,
0.301543802,
0.333420366,
-0.0705919489,
-0.4270... |
http://arxiv.org/abs/2406.14887v1 | http://arxiv.org/pdf/2406.14887v1 | InternLM-Law: An Open Source Chinese Legal Large Language Model | 2024-06-21T06:19:03 | 2024-06-21T06:19:03 | Zhiwei Fei, Songyang Zhang, Xiaoyu Shen, Dawei Zhu, Xiao Wang, Maosong Cao, Fengzhe Zhou, Yining Li, Wenwei Zhang, Dahua Lin, Kai Chen, Jidong Ge | cs.CL | cs.CL | While large language models (LLMs) have showcased impressive capabilities,
they struggle with addressing legal queries due to the intricate complexities
and specialized expertise required in the legal field. In this paper, we
introduce InternLM-Law, a specialized LLM tailored for addressing diverse legal
queries relate... | [
-0.0876411274,
-0.5878869295,
0.2383494228,
0.301733762,
-0.5512210131,
-0.1970692873,
-0.4774421453,
-0.4461461306,
-0.301291436,
0.7263495922,
-0.1735903174,
-0.7450985909,
-0.2211245149,
-0.2289007604,
-0.6434304118,
1.0165826082,
0.0998450667,
0.597577095,
0.1060827672,
-0.... |
http://arxiv.org/abs/2406.14884v1 | http://arxiv.org/pdf/2406.14884v1 | FlowBench: Revisiting and Benchmarking Workflow-Guided Planning for LLM-based Agents | 2024-06-21T06:13:00 | 2024-06-21T06:13:00 | Ruixuan Xiao, Wentao Ma, Ke Wang, Yuchuan Wu, Junbo Zhao, Haobo Wang, Fei Huang, Yongbin Li | cs.CL | cs.CL | LLM-based agents have emerged as promising tools, which are crafted to
fulfill complex tasks by iterative planning and action. However, these agents
are susceptible to undesired planning hallucinations when lacking specific
knowledge for expertise-intensive tasks. To address this, preliminary attempts
are made to enhan... | [
-0.3661183119,
-0.2917479873,
0.589648068,
0.114890486,
-0.0188579187,
0.2298982888,
0.0963403508,
-0.3052945435,
0.3380846083,
0.6127391458,
-0.8729587197,
-0.4039192796,
-0.4676786661,
-0.076964803,
-0.2660242319,
0.5969730616,
-0.4251089096,
0.1842769235,
-0.2627724707,
-0.2... |
http://arxiv.org/abs/2406.14883v1 | http://arxiv.org/pdf/2406.14883v1 | OATH-Frames: Characterizing Online Attitudes Towards Homelessness with LLM Assistants | 2024-06-21T06:09:47 | 2024-06-21T06:09:47 | Jaspreet Ranjit, Brihi Joshi, Rebecca Dorn, Laura Petry, Olga Koumoundouros, Jayne Bottarini, Peichen Liu, Eric Rice, Swabha Swayamdipta | cs.CL | cs.CL, cs.CY | Warning: Contents of this paper may be upsetting.
Public attitudes towards key societal issues, expressed on online media, are
of immense value in policy and reform efforts, yet challenging to understand at
scale. We study one such social issue: homelessness in the U.S., by leveraging
the remarkable capabilities of l... | [
-0.2014110982,
-0.5827611685,
0.6715899706,
0.3854842782,
0.0431597978,
-0.0433329344,
0.0294949245,
-0.3789282441,
0.6312223077,
0.0357381776,
-0.1963496506,
-0.1552781165,
-1.0132079124,
0.2911745608,
-0.327154398,
0.9161157012,
0.0304171164,
0.2683514953,
-0.3361695409,
-0.3... |
http://arxiv.org/abs/2406.14882v1 | http://arxiv.org/pdf/2406.14882v1 | 70B-parameter large language models in Japanese medical question-answering | 2024-06-21T06:04:10 | 2024-06-21T06:04:10 | Issey Sukeda, Risa Kishikawa, Satoshi Kodera | cs.CL | cs.CL | Since the rise of large language models (LLMs), the domain adaptation has
been one of the hot topics in various domains. Many medical LLMs trained with
English medical dataset have made public recently. However, Japanese LLMs in
medical domain still lack its research. Here we utilize multiple 70B-parameter
LLMs for the... | [
-0.3242919147,
-0.9532623887,
0.7706362009,
-0.1558904052,
-0.3346259594,
-0.112415053,
-0.4226923585,
-0.2322345078,
0.0109470896,
0.6722555161,
-0.5904121399,
-0.7145259976,
-0.5888057351,
0.2666176558,
0.3153199553,
0.8162095547,
-0.1975421757,
0.6581301689,
-0.343483001,
-0... |
http://arxiv.org/abs/2406.14877v1 | http://arxiv.org/pdf/2406.14877v1 | Sports Intelligence: Assessing the Sports Understanding Capabilities of Language Models through Question Answering from Text to Video | 2024-06-21T05:57:50 | 2024-06-21T05:57:50 | Zhengbang Yang, Haotian Xia, Jingxi Li, Zezhi Chen, Zhuangdi Zhu, Weining Shen | cs.CL | cs.CL | Understanding sports is crucial for the advancement of Natural Language
Processing (NLP) due to its intricate and dynamic nature. Reasoning over
complex sports scenarios has posed significant challenges to current NLP
technologies which require advanced cognitive capabilities. Toward addressing
the limitations of exist... | [
-0.2451942265,
-0.872741282,
0.5589591861,
0.0425388664,
-0.1136154085,
0.1630865335,
-0.3587570488,
-0.636730969,
-0.2666652203,
0.1266685575,
-0.9116162658,
-0.4761208296,
-0.7109377384,
-0.021016581,
-0.2851650715,
0.8604739308,
-0.2225282341,
0.1971439123,
-0.5491365194,
-0... |
http://arxiv.org/abs/2406.15524v1 | http://arxiv.org/pdf/2406.15524v1 | Rethinking Pruning Large Language Models: Benefits and Pitfalls of Reconstruction Error Minimization | 2024-06-21T05:13:34 | 2024-06-21T05:13:34 | Sungbin Shin, Wonpyo Park, Jaeho Lee, Namhoon Lee | cs.CL | cs.CL, cs.LG | This work suggests fundamentally rethinking the current practice of pruning
large language models (LLMs). The way it is done is by divide and conquer:
split the model into submodels, sequentially prune them, and reconstruct
predictions of the dense counterparts on small calibration data one at a time;
the final model i... | [
-0.2900706828,
-0.6390492916,
0.5001763105,
0.0287378896,
-0.160756886,
0.1269646138,
-0.317707926,
-0.1092837974,
0.2432634532,
0.7519919872,
-0.5065328479,
-0.17888394,
-0.489115268,
-0.0686080232,
-0.3676677048,
0.7697650194,
0.1060878262,
-0.1429995596,
-0.2218632698,
-0.00... |
http://arxiv.org/abs/2406.14868v2 | http://arxiv.org/pdf/2406.14868v2 | Direct Multi-Turn Preference Optimization for Language Agents | 2024-06-21T05:13:20 | 2024-06-25T08:44:24 | Wentao Shi, Mengqi Yuan, Junkang Wu, Qifan Wang, Fuli Feng | cs.CL | cs.CL, cs.LG | Adapting Large Language Models (LLMs) for agent tasks is critical in
developing language agents. Direct Preference Optimization (DPO) is a promising
technique for this adaptation with the alleviation of compounding errors,
offering a means to directly optimize Reinforcement Learning (RL) objectives.
However, applying D... | [
-0.2617801428,
-0.2238533348,
0.9621323347,
0.4066695869,
-0.0749967694,
0.1234806776,
-0.5188307762,
0.1314683557,
-0.0103785582,
0.6480264664,
-0.8491196632,
-0.272657603,
-0.5114132166,
0.0808659643,
0.2402634323,
0.9627546668,
-0.4027344882,
-0.1956750154,
-0.1134094819,
-0... |
http://arxiv.org/abs/2406.14867v1 | http://arxiv.org/pdf/2406.14867v1 | DistiLRR: Transferring Code Repair for Low-Resource Programming Languages | 2024-06-21T05:05:39 | 2024-06-21T05:05:39 | Kyle Wong, Alfonso Amayuelas, Liangming Pan, William Yang Wang | cs.LG | cs.LG, cs.AI, cs.CL | Large language models (LLMs) have shown remarkable performance on code
generation tasks. A recent application of LLMs for code generation is iterative
code repair, where a model fixes an incorrect program by rationalizing about
errors and generating a new program. However, code repair is primarily studied
on high-resou... | [
-0.0426562838,
-0.4990505576,
0.3738858104,
0.1217652708,
-0.0112273199,
0.4328177571,
-0.4278483391,
-0.1812734455,
-0.1830942631,
0.4566753507,
-0.2623055577,
0.1210589558,
-0.536585629,
0.1542649418,
-0.3207958341,
1.1498318911,
-0.05556041,
-0.0090446025,
-0.4046064317,
-0.... |
http://arxiv.org/abs/2406.14862v3 | http://arxiv.org/pdf/2406.14862v3 | LatentExplainer: Explaining Latent Representations in Deep Generative Models with Multi-modal Foundation Models | 2024-06-21T04:39:03 | 2024-06-28T13:19:37 | Mengdan Zhu, Raasikh Kanjiani, Jiahui Lu, Andrew Choi, Qirui Ye, Liang Zhao | cs.LG | cs.LG, cs.CL, cs.CV | Deep generative models like VAEs and diffusion models have advanced various
generation tasks by leveraging latent variables to learn data distributions and
generate high-quality samples. Despite the field of explainable AI making
strides in interpreting machine learning models, understanding latent variables
in generat... | [
-0.1988127679,
-0.7469796538,
0.647277534,
0.33406201,
-0.2031324953,
0.0848316848,
0.0086672641,
0.111353986,
-0.249080047,
0.2267169803,
-0.7826244831,
-0.3512734473,
-0.3487448096,
-0.2436967045,
-0.2000176013,
0.9030354023,
-0.254784584,
0.0436593816,
-0.1917851865,
0.23900... |
http://arxiv.org/abs/2406.14859v1 | http://arxiv.org/pdf/2406.14859v1 | From LLMs to MLLMs: Exploring the Landscape of Multimodal Jailbreaking | 2024-06-21T04:33:48 | 2024-06-21T04:33:48 | Siyuan Wang, Zhuohan Long, Zhihao Fan, Zhongyu Wei | cs.CL | cs.CL, cs.AI | The rapid development of Large Language Models (LLMs) and Multimodal Large
Language Models (MLLMs) has exposed vulnerabilities to various adversarial
attacks. This paper provides a comprehensive overview of jailbreaking research
targeting both LLMs and MLLMs, highlighting recent advancements in evaluation
benchmarks, a... | [
-0.3913839757,
-1.103420496,
0.5488118529,
0.1760519147,
0.1771147549,
0.7152983546,
-0.2076319307,
-0.4566475451,
0.0441955887,
0.7275462151,
-0.6384664178,
-0.2285315543,
-0.7917206883,
-0.0974353179,
-0.3440654874,
1.0203640461,
-0.0085213212,
0.3465019763,
-0.2306685597,
-0... |
http://arxiv.org/abs/2406.14848v1 | http://arxiv.org/pdf/2406.14848v1 | Leveraging Passage Embeddings for Efficient Listwise Reranking with Large Language Models | 2024-06-21T03:33:51 | 2024-06-21T03:33:51 | Qi Liu, Bo Wang, Nan Wang, Jiaxin Mao | cs.CL | cs.CL, cs.IR | Recent studies have demonstrated the effectiveness of using large language
language models (LLMs) in passage ranking. The listwise approaches, such as
RankGPT, have become new state-of-the-art in this task. However, the efficiency
of RankGPT models is limited by the maximum context length and relatively high
latency of... | [
-0.0620161183,
-0.5265398622,
0.7161983848,
0.2828380167,
-0.5339554548,
0.2173992395,
-0.2514661849,
0.1025108844,
0.3134961426,
0.4018638134,
-0.0623109639,
-0.6501509547,
-0.8608752489,
0.1492316127,
-0.2802045345,
0.8796911836,
0.2443444729,
-0.1666274965,
-0.0581035726,
0.... |
http://arxiv.org/abs/2406.14835v1 | http://arxiv.org/pdf/2406.14835v1 | ToVo: Toxicity Taxonomy via Voting | 2024-06-21T02:35:30 | 2024-06-21T02:35:30 | Tinh Son Luong, Thanh-Thien Le, Thang Viet Doan, Linh Ngo Van, Thien Huu Nguyen, Diep Thi-Ngoc Nguyen | cs.CL | cs.CL, cs.LG | Existing toxic detection models face significant limitations, such as lack of
transparency, customization, and reproducibility. These challenges stem from
the closed-source nature of their training data and the paucity of explanations
for their evaluation mechanism. To address these issues, we propose a dataset
creatio... | [
-0.0762640089,
-0.1411909908,
0.7002923489,
-0.110541068,
-0.4855606258,
-0.4038804471,
0.0020062993,
-0.0332203843,
0.0307380427,
0.8181284666,
-0.5004994869,
-1.2267867327,
-0.4899691641,
-0.0818408653,
-0.3153610229,
1.2392910719,
0.1435873359,
-0.3063006699,
-0.2877734601,
... |
http://arxiv.org/abs/2406.14833v2 | http://arxiv.org/pdf/2406.14833v2 | Efficient Continual Pre-training by Mitigating the Stability Gap | 2024-06-21T02:28:37 | 2024-06-27T08:11:01 | Yiduo Guo, Jie Fu, Huishuai Zhang, Dongyan Zhao, Yikang Shen | cs.CL | cs.CL | Continual pre-training has increasingly become the predominant approach for
adapting Large Language Models (LLMs) to new domains. This process involves
updating the pre-trained LLM with a corpus from a new domain, resulting in a
shift in the training distribution. To study the behavior of LLMs during this
shift, we mea... | [
-0.3078753352,
-0.6359675527,
0.5658908486,
0.0657630265,
-0.4120950103,
0.0796901286,
-0.3848050237,
-0.4116501212,
0.0502348505,
0.5205618143,
-0.4375278652,
-0.4490173757,
-0.8047896028,
0.0620979406,
-0.1720563769,
0.7583017945,
-0.1182920933,
0.1838915944,
-0.1713800132,
-... |
http://arxiv.org/abs/2406.14829v1 | http://arxiv.org/pdf/2406.14829v1 | Is this a bad table? A Closer Look at the Evaluation of Table Generation from Text | 2024-06-21T02:18:03 | 2024-06-21T02:18:03 | Pritika Ramu, Aparna Garimella, Sambaran Bandyopadhyay | cs.CL | cs.CL | Understanding whether a generated table is of good quality is important to be
able to use it in creating or editing documents using automatic methods. In
this work, we underline that existing measures for table quality evaluation
fail to capture the overall semantics of the tables, and sometimes unfairly
penalize good ... | [
-0.3146258593,
-1.0187363625,
0.362424612,
-0.2335915118,
-0.4743891954,
-0.0607451759,
0.0094467094,
-0.2587066591,
0.1266553402,
0.5436454415,
-0.1399129331,
-0.7939656377,
-0.260479033,
0.1730964184,
-0.8994719386,
1.5875356197,
-0.1244962141,
-0.1901410669,
-0.5979304314,
0... |
http://arxiv.org/abs/2406.14828v1 | http://arxiv.org/pdf/2406.14828v1 | Word Matters: What Influences Domain Adaptation in Summarization? | 2024-06-21T02:15:49 | 2024-06-21T02:15:49 | Yinghao Li, Siyu Miao, Heyan Huang, Yang Gao | cs.CL | cs.CL | Domain adaptation aims to enable Large Language Models (LLMs) to generalize
domain datasets unseen effectively during the training phase. However, factors
such as the size of the model parameters and the scale of training data are
general influencers and do not reflect the nuances of domain adaptation
performance. This... | [
-0.4382908046,
-0.6260759234,
0.3946126401,
0.2428129315,
-0.3401474059,
0.1327447742,
-0.3334878981,
-0.4720717967,
-0.1568279862,
0.1360862106,
-0.4061954916,
-0.149115324,
-1.316640377,
0.269939214,
-0.103337355,
1.2517861128,
-0.1676586419,
-0.092720367,
-0.5425152779,
-0.1... |
http://arxiv.org/abs/2406.14825v3 | http://arxiv.org/pdf/2406.14825v3 | TemPrompt: Multi-Task Prompt Learning for Temporal Relation Extraction in RAG-based Crowdsourcing Systems | 2024-06-21T01:52:37 | 2024-06-30T08:54:37 | Jing Yang, Yu Zhao, Yang Linyao, Xiao Wang, Long Chen, Fei-Yue Wang | cs.CL | cs.CL | Temporal relation extraction (TRE) aims to grasp the evolution of events or
actions, and thus shape the workflow of associated tasks, so it holds promise
in helping understand task requests initiated by requesters in crowdsourcing
systems. However, existing methods still struggle with limited and unevenly
distributed a... | [
-0.3227316439,
-0.4544782937,
0.5908844471,
0.2633718848,
-0.5615876913,
0.1822226793,
-0.3084515929,
-0.3967092633,
-0.0222025607,
0.0960945562,
-0.8690379858,
-0.4756979346,
-0.4720302224,
0.3040677309,
0.0803912058,
1.0172860622,
0.1188946515,
-0.1730765253,
-0.0367315449,
-... |
http://arxiv.org/abs/2406.15518v1 | http://arxiv.org/pdf/2406.15518v1 | Steering Without Side Effects: Improving Post-Deployment Control of Language Models | 2024-06-21T01:37:39 | 2024-06-21T01:37:39 | Asa Cooper Stickland, Alexander Lyzhov, Jacob Pfau, Salsabila Mahdi, Samuel R. Bowman | cs.CL | cs.CL, cs.LG | Language models (LMs) have been shown to behave unexpectedly post-deployment.
For example, new jailbreaks continually arise, allowing model misuse, despite
extensive red-teaming and adversarial training from developers. Given most
model queries are unproblematic and frequent retraining results in unstable
user experien... | [
-0.5308645368,
-0.5779099464,
0.3876721263,
0.1934012324,
-0.332713604,
0.2511671782,
-0.3085749149,
-0.196479544,
0.144757241,
0.7470179796,
-0.4088102281,
-0.4465480447,
-0.6678091884,
-0.4752391875,
-0.5765707493,
1.5123556852,
0.1780601591,
-0.1111608669,
-0.1435254365,
-0.... |
http://arxiv.org/abs/2406.14805v1 | http://arxiv.org/pdf/2406.14805v1 | How Well Do LLMs Represent Values Across Cultures? Empirical Analysis of LLM Responses Based on Hofstede Cultural Dimensions | 2024-06-21T00:58:01 | 2024-06-21T00:58:01 | Julia Kharchenko, Tanya Roosta, Aman Chadha, Chirag Shah | cs.CL | cs.CL | Large Language Models (LLMs) attempt to imitate human behavior by responding
to humans in a way that pleases them, including by adhering to their values.
However, humans come from diverse cultures with different values. It is
critical to understand whether LLMs showcase different values to the user based
on the stereot... | [
-0.2408407331,
-0.5290918946,
0.4794009626,
0.54668957,
-0.3407595158,
0.3532068729,
-0.2169512808,
-0.6952511072,
0.4267140031,
0.4213166833,
-0.4987220764,
-0.4653202295,
-0.6373121142,
0.1047497168,
0.0706351027,
0.8613319397,
-0.0496655777,
0.4194640517,
-0.4443563521,
-0.5... |
http://arxiv.org/abs/2406.14785v1 | http://arxiv.org/pdf/2406.14785v1 | Understanding Finetuning for Factual Knowledge Extraction | 2024-06-20T23:27:06 | 2024-06-20T23:27:06 | Gaurav Ghosal, Tatsunori Hashimoto, Aditi Raghunathan | cs.CL | cs.CL, cs.LG | In this work, we study the impact of QA fine-tuning data on downstream
factuality. We show that fine-tuning on lesser-known facts that are poorly
stored during pretraining yields significantly worse factuality than
fine-tuning on well-known facts, even when all facts are seen during
pretraining. We prove this phenomeno... | [
-0.6569651961,
-0.8682034612,
0.866761446,
-0.4550442696,
-0.3811980784,
-0.3634667099,
-0.352951169,
-0.5869572759,
-0.3327288926,
0.4917747676,
-0.6359158158,
-0.2605485022,
-0.6256270409,
-0.0472220369,
-0.3762564957,
1.460261941,
0.2379077822,
0.1144069433,
-0.328235805,
-0... |
http://arxiv.org/abs/2406.14783v1 | http://arxiv.org/pdf/2406.14783v1 | Evaluating RAG-Fusion with RAGElo: an Automated Elo-based Framework | 2024-06-20T23:20:34 | 2024-06-20T23:20:34 | Zackary Rackauckas, Arthur Câmara, Jakub Zavrel | cs.IR | cs.IR, cs.CL | Challenges in the automated evaluation of Retrieval-Augmented Generation
(RAG) Question-Answering (QA) systems include hallucination problems in
domain-specific knowledge and the lack of gold standard benchmarks for company
internal tasks. This results in difficulties in evaluating RAG variations, like
RAG-Fusion (RAGF... | [
-0.5586578846,
-0.4158605635,
0.5159525275,
0.1188773662,
0.0896148682,
0.5193481445,
-0.1138097197,
-0.7184799314,
0.1897474378,
0.45463866,
-0.3247149587,
-0.5353314281,
-0.4203086793,
0.4810196459,
-0.5591688156,
1.255135417,
-0.0617282279,
-0.0844174251,
-0.7393070459,
-0.1... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.