benchmark_id stringlengths 12 56 | slug stringlengths 3 56 | name stringlengths 2 44 | source stringclasses 4
values | source_url stringlengths 19 110 | description stringlengths 0 995 | categories listlengths 0 21 | languages listlengths 0 13 | modality stringclasses 5
values | publisher stringlengths 3 80 ⌀ | released_at timestamp[s]date 2010-02-19 00:00:00 2026-08-25 00:00:00 ⌀ | openness stringclasses 3
values | paper_url stringlengths 31 62 ⌀ | repo_url stringlengths 28 89 ⌀ | dataset_url stringlengths 40 81 ⌀ | document_count int64 1 27 ⌀ | model_count int64 1 586 ⌀ | score_count int64 0 586 | highest_score float64 0.01 2.1M ⌀ | score_unit stringclasses 3
values |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
llm-stats:attaq | llm-stats-attaq | AttaQ | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/attaq?top_n=500 | AttaQ is a unique dataset containing adversarial examples in the form of questions designed to provoke harmful or inappropriate responses from large language models. The benchmark evaluates safety vulnerabilities by using specialized clustering techniques that analyze both the semantic similarity of input attacks and t... | [
"safety"
] | [] | text | null | null | unknown | null | null | null | 1 | 3 | 3 | 0.885 | null |
opencompass:1835 | opencompass-1835-audiojailbreak | AudioJailbreak | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/AudioJailbreak | LAMs face jailbreak risks. AJailBench, our new benchmark, reveals leading LAMs lack robustness. Subtle audio perturbations significantly degrade their safety. We release AJailBench for research. LAM 面临越狱风险。我们新的基准测试 AJailBench 揭示，领先的 LAM 缺乏稳健性。细微的音频干扰会显著降低其安全性。我们发布 AJailBench 进行研究。 | [
"多模态",
"Multimodal",
"语言",
"Language",
"安全",
"Safety",
"audio",
"jailbreak",
"LAM",
"多模态模型",
"VLM",
"语言理解",
"Comprehension",
"安全对齐",
"Safety Alignment",
"音频理解",
"Audio Understanding",
"不支持",
"Unsupported"
] | [] | multimodal | mbzuai | 2025-05-24T00:00:00 | open | https://arxiv.org/abs/2505.15406 | https://github.com/mbzuai-nlp/AudioJailbreak | https://huggingface.co/datasets/MBZUAI/AudioJailbreak | 1 | null | 0 | null | null |
opencompass:1908 | opencompass-1908-audiotrust | AudioTrust | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/AudioTrust | AudioTrust is a comprehensive trust evaluation framework for Audio Large Language Models (ALLMs) that effectively reveals potential risks in six dimensions: fairness, hallucination, security, privacy, robustness, and authentication. It aggregates over 4,420 real-world audio/text data samples, coveri AudioTrust针对Audio L... | [
"多模态",
"Multimodal",
"安全",
"Safety",
"多模态模型",
"VLM",
"安全对齐",
"Safety Alignment",
"音频理解",
"Audio Understanding",
"不支持",
"Unsupported"
] | [
"English"
] | multimodal | Tsinghua University, Nanyang Technological University | 2025-06-06T00:00:00 | open | https://arxiv.org/abs/2505.16211 | https://github.com/JusperLee/AudioTrust | https://huggingface.co/datasets/JusperLee/AudioTrust | 1 | null | 0 | null | null |
opencompass:2078 | opencompass-2078-autoadvexbench | AutoAdvExBench | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/AutoAdvExBench | AutoAdvExBench is a benchmark designed to evaluate large language models' (LLMs) ability to autonomously exploit adversarial example defenses, directly measuring LLMs' success on tasks regularly performed by machine learning security experts. AutoAdvExBench 是一个评估大型语言模型（LLMs）自主利用对抗性样本防御能力的基准，直接衡量LLMs在机器学习安全专家任务上的成功率。它主要... | [
"安全",
"Safety",
"智能体",
"Agent",
"任务执行",
"Task Execution",
"安全对齐",
"Safety Alignment",
"不支持",
"Unsupported"
] | [] | null | GoogleDeepMind , ETHZurich | 2025-03-03T00:00:00 | unknown | https://arxiv.org/abs/2503.01811 | https://github.com/ethz-spylab/AutoAdvExBench | null | 1 | null | 0 | null | null |
llm-stats:autologi | llm-stats-autologi | AutoLogi | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/autologi?top_n=500 | AutoLogi is an automated method for synthesizing open-ended logic puzzles to evaluate reasoning abilities of Large Language Models. The benchmark addresses limitations of existing multiple-choice reasoning evaluations by featuring program-based verification and controllable difficulty levels. It includes 1,575 English ... | [
"reasoning"
] | [] | text | null | null | unknown | null | null | null | 1 | 2 | 2 | 0.895 | null |
llm-stats:automationbench | llm-stats-automationbench | AutomationBench | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/automationbench?top_n=500 | AutomationBench is a tool-use benchmark that evaluates AI agents on automating real-world workflows, testing their ability to orchestrate tools and complete multi-step automation tasks. | [
"reasoning",
"agents",
"tool_calling"
] | [] | text | null | null | unknown | null | null | null | 1 | 12 | 12 | 0.482 | null |
model-reports:automationbench | automationbench | AutomationBench | model_reports | https://github.com/zapier/automation-bench | Published by Zapier over its own automation surface, and cards report different task subsets of it. | [
"agent"
] | [] | null | null | 2026-03-10T00:00:00 | unknown | null | https://github.com/zapier/automation-bench | null | 6 | 5 | 5 | 48.8 | percent |
artificial-analysis:automationbench-aa | artificial-analysis-automationbench-aa | AutomationBench-AA | artificial_analysis | https://artificialanalysis.ai/evaluations/automationbench-aa | Agentic SaaS workflows | [
"agentic",
"tool-use",
"business"
] | [] | null | null | null | unknown | null | null | null | 1 | 39 | 39 | 0.627456 | null |
llm-stats:automationbench-aa | llm-stats-automationbench-aa | AutomationBench-AA | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/automationbench-aa?top_n=500 | AutomationBench-AA is Artificial Analysis's independently run version of AutomationBench, covering 657 real-world SaaS workflow tasks across 40 simulated applications (e.g. Gmail, Slack, Salesforce, HubSpot). It scores the share of objectives an agent completes without violating business guardrails, using a private hel... | [
"reasoning",
"agents",
"tool_calling"
] | [] | text | null | null | unknown | null | null | null | 1 | 1 | 1 | 0.514 | null |
opencompass:1396 | opencompass-1396-av-odyssey-bench | AV-Odyssey-Bench | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/AV-Odyssey-Bench | AV-Odyssey Bench. This benchmark encompasses 26 different tasks and 4,555 carefully crafted problems, each incorporating text, visual, and audio components. All data are newly collected and annotated by humans, not from any existing audio-visual dataset. AV-Odyssey Bench. This benchmark encompasses 26 different tasks a... | [
"多模态",
"Multimodal",
"其他",
"Other",
"audio-visual",
"多模态模型",
"VLM",
"音频理解",
"Audio Understanding",
"不支持",
"Unsupported"
] | [] | multimodal | CUHK MMLab, CUHK (SZ), Stanford University, UC Berkeley, Yale University | 2024-12-03T00:00:00 | open | https://arxiv.org/pdf/2412.02611 | https://github.com/AV-Odyssey/AV-Odyssey | https://huggingface.co/datasets/AV-Odyssey/AV_Odyssey_Bench | 1 | null | 0 | null | null |
opencompass:526 | opencompass-526-ax-b | AX-b | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/AX-b | AX-b is a broad-coverage diagnostic task, which requires to determine the logical relation between the given sentence pair, with three relations: entailment, contradiction and neutral. This task is selected from a subset of the GLUE broad-coverage diagnostic dataset, mainly used to test the model's understanding abilit... | [
"推理",
"Reasoning",
"大语言模型",
"LLM",
"逻辑推理",
"开源收录",
"Open-Source",
"不支持",
"Unsupported"
] | [] | null | null | 2019-05-02T00:00:00 | unknown | https://arxiv.org/abs/1905.00537 | null | null | 1 | null | 0 | null | null |
opencompass:527 | opencompass-527-ax-g | AX-g | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/AX-g | AX-g is a Winogender diagnostic task, which requires to determine which noun the pronoun refers to according to the given sentence and pronoun. This task is selected from a subset of the Winogender dataset, mainly used to test the model's ability in dealing with gender bias and discrimination. AX-g是一个Winogender诊断任务，要求根... | [
"推理",
"Reasoning",
"大语言模型",
"LLM",
"逻辑推理",
"开源收录",
"Open-Source",
"不支持",
"Unsupported"
] | [] | null | null | 2019-07-12T00:00:00 | unknown | https://arxiv.org/abs/1905.00537 | https://github.com/rudinger/winogender-schemas | null | 1 | null | 0 | null | null |
opencompass:2084 | opencompass-2084-axbench | AXBENCH | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/AXBENCH | AXBENCH is a benchmark for large-scale evaluation of Language Model (LLM) control methods using synthetic data, focusing on fine-grained steering for safety and reliability. AXBENCH 是一个旨在评估LLM控制能力的基准。它通过概念检测和模型操控（包含概念、指令、流畅度）评估，旨在实现安全可靠的细粒度操控。基准使用大规模合成数据集，并集成了多种基线方法。 | [
"理解",
"Understanding",
"大语言模型",
"LLM",
"语言理解",
"Comprehension",
"不支持",
"Unsupported"
] | [] | null | Department of Computer Science,Stanford University , Pr(AI) R Group. | 2025-01-28T00:00:00 | unknown | https://arxiv.org/abs/2501.17148 | https://github.com/stanfordnlp/axbench | null | 1 | null | 0 | null | null |
opencompass:1266 | opencompass-1266-babilong | BABILong | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/BABILong | BABILong is designed to test language models' ability to reason across facts distributed in extremely long documents. It contains a diverse set of 20 reasoning tasks, including fact chaining, simple induction, deduction, counting, and handling lists/sets. BABILong旨在测试语言模型对分布在极长文档中的事实进行推理的能力，涵盖事实链接、简单归纳、推导、计数和处理列表/集合等20... | [
"长文本",
"Long-Context",
"NeurIPS 2024",
"大语言模型",
"LLM",
"长上下文",
"Long Context",
"开源收录",
"Open-Source",
"不支持",
"Unsupported"
] | [] | null | AIRI | 2024-06-14T00:00:00 | restricted | https://arxiv.org/abs/1502.05698 | https://github.com/booydar/babilong | https://huggingface.co/datasets/RMT-team/babilong | 1 | null | 0 | null | null |
llm-stats:babyvision | llm-stats-babyvision | BabyVision | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/babyvision?top_n=500 | A benchmark for early-stage visual reasoning and perception on child-like vision tasks. | [
"multimodal",
"reasoning",
"vision"
] | [] | multimodal | null | null | unknown | null | null | null | 1 | 10 | 10 | 0.857 | null |
model-reports:babyvision | babyvision | BabyVision | model_reports | https://github.com/babyvision/babyvision | Multimodal vision benchmark. Temperature=1.0, top_p=0.95, max context 164K tokens. Images resized to shorter side at least 1.5K pixels. | [
"multimodal"
] | [] | null | null | 2025-12-01T00:00:00 | unknown | null | https://github.com/babyvision/babyvision | null | 1 | 1 | 1 | 53.4 | percent |
llm-stats:bankertoolbench | llm-stats-bankertoolbench | BankerToolBench | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/bankertoolbench?top_n=500 | BankerToolBench is a public benchmark that evaluates models on banking and finance tool-use tasks. Models are scored against dataset rubrics, measuring their ability to correctly invoke tools and complete multi-step financial workflows. | [
"finance",
"agents",
"tool_calling"
] | [] | text | null | null | unknown | null | null | null | 1 | 1 | 1 | 0.7612 | null |
model-reports:bankertoolbench | bankertoolbench | BankerToolBench | model_reports | https://github.com/Handshake-AI-Research/bankertoolbench | End-to-end investment-banking tasks produce spreadsheets, presentations, and documents; the agent harness, financial-data tools, and rubric-grader configuration are part of the score. | [
"professional"
] | [] | null | null | null | unknown | null | https://github.com/Handshake-AI-Research/bankertoolbench | null | 1 | 1 | 1 | 78.6 | percent |
llm-stats:bbh | llm-stats-bbh | BBH | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/bbh?top_n=500 | Big-Bench Hard (BBH) is a suite of 23 challenging tasks selected from BIG-Bench for which prior language model evaluations did not outperform the average human-rater. These tasks require multi-step reasoning across diverse domains including arithmetic, logical reasoning, reading comprehension, and commonsense reasoning... | [
"math",
"reasoning",
"language"
] | [] | text | null | null | unknown | null | null | null | 1 | 12 | 12 | 0.8887 | null |
opencompass:539 | opencompass-539-bbh | BBH | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/BBH | BIG-Bench Hard (BBH) is a subset of the BIG-Bench, a diverse evaluation suite for language models. BBH focuses on a suite of 23 challenging tasks from BIG-Bench that were found to be beyond the capabilities of current language models. BIG Bench-Hard（BBH）是BIG Bench的一个子集，它是一个用于语言模型的多样化评估套件。BBH专注于BIG Bench的23项具有挑战性的任务，这些任... | [
"推理",
"Reasoning",
"大语言模型",
"LLM",
"逻辑推理",
"开源收录",
"Open-Source",
"支持",
"Supported"
] | [] | null | null | 2022-10-17T00:00:00 | unknown | https://arxiv.org/pdf/2210.09261.pdf | https://github.com/suzgunmirac/BIG-Bench-Hard | null | 1 | null | 0 | null | null |
llm-stats:bc-vl | llm-stats-bc-vl | BC-VL | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/bc-vl?top_n=500 | BC-VL is a vision-language benchmark for knowledge-grounded multimodal question answering. | [
"multimodal",
"knowledge",
"vision"
] | [] | multimodal | null | null | unknown | null | null | null | 1 | 1 | 1 | 0.511 | null |
llm-stats:beam-128k | llm-stats-beam-128k | Beam 128K | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/beam-128k?top_n=500 | Beam 128K evaluates reasoning over long inputs at a 128K-token context length. | [
"long_context",
"reasoning"
] | [] | text | null | null | unknown | null | null | null | 1 | 1 | 1 | 0.651 | null |
opencompass:1086 | opencompass-1086-belebele | Belebele | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/Belebele | BELEBELE, a multiple-choice machine reading comprehension (MRC) dataset spanning 122 language variants. Significantly expanding the language coverage of natural language understanding (NLU) benchmarks, this dataset enables the evaluation of text models in
high-, medium-, and low-resource languages. BELEBELE 是一个多项选择机器阅读... | [
"语言",
"Language",
"大语言模型",
"LLM",
"语言理解",
"Comprehension",
"不支持",
"Unsupported"
] | [] | null | FaceBook | 2024-07-25T00:00:00 | unknown | https://arxiv.org/pdf/2308.16884 | https://github.com/facebookresearch/belebele | null | 1 | null | 0 | null | null |
llm-stats:benchcad | llm-stats-benchcad | BenchCAD | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/benchcad?top_n=500 | BenchCAD is a benchmark for programmatic CAD reasoning built from 17,900 execution-verified CadQuery programs spanning 106 industrial part families, roughly half anchored to real ISO, DIN, EN, ASME, and IEC specification tables. It decomposes CAD capability into matched tasks; the Vision2Code task requires models to ge... | [
"multimodal",
"reasoning",
"code",
"vision"
] | [] | multimodal | null | null | unknown | null | null | null | 1 | 4 | 4 | 0.706 | null |
llm-stats:benchcad-with-python-tool | llm-stats-benchcad-with-python-tool | BenchCAD (with Python tool) | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/benchcad-with-python-tool?top_n=500 | BenchCAD variant evaluated with access to a Python tool for programmatic CAD reasoning. | [
"multimodal",
"reasoning",
"code",
"vision"
] | [] | multimodal | null | null | unknown | null | null | null | 1 | 3 | 3 | 0.834 | null |
opencompass:1562 | opencompass-1562-benchmax | BenchMAX | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/BenchMAX | BenchMAX is a comprehensive, high-quality, and multiway parallel multilingual benchmark comprising 10 tasks designed to assess crucial capabilities across 17 diverse language. BenchMAX 是一个全面、高质量的多向并行多语言基准，包含 10 个任务，旨在评估 17 种不同语言的关键能力。 | [
"语言",
"Language",
"大语言模型",
"LLM",
"语言理解",
"Comprehension",
"不支持",
"Unsupported"
] | [
"English",
"Chinese",
"Japanese",
"Korean",
"French",
"German",
"Spanish",
"Arabic",
"Russian",
"Vietnamese",
"Thai",
"Multilingual"
] | null | National Key Laboratory for Novel Software Technology, Nanjing University, etc. | 2025-02-11T00:00:00 | unknown | https://arxiv.org/abs/2502.07346 | https://github.com/CONE-MT/BenchMAX | null | 1 | null | 0 | null | null |
llm-stats:beyond-aime | llm-stats-beyond-aime | Beyond AIME | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/beyond-aime?top_n=500 | Beyond AIME is a difficult mathematical reasoning benchmark designed to test deeper reasoning chains and harder decomposition than standard AIME-style problem sets. | [
"math",
"reasoning"
] | [] | text | null | null | unknown | null | null | null | 1 | 5 | 5 | 0.88 | null |
llm-stats:bfcl | llm-stats-bfcl | BFCL | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/bfcl?top_n=500 | The Berkeley Function Calling Leaderboard (BFCL) is the first comprehensive and executable function call evaluation dedicated to assessing Large Language Models' ability to invoke functions. It evaluates serial and parallel function calls across multiple programming languages (Python, Java, JavaScript, REST API) using ... | [
"reasoning",
"general",
"tool_calling"
] | [] | text | null | null | unknown | null | null | null | 1 | 11 | 11 | 0.885 | null |
model-reports:bfcl | bfcl | BFCL | model_reports | https://gorilla.cs.berkeley.edu/leaderboard.html | Schema complexity and execution checking vary by version. | [
"tool_use"
] | [] | null | null | 2024-02-26T00:00:00 | unknown | null | null | null | 4 | 2 | 2 | 72.9 | percent |
llm-stats:bfcl-v2 | llm-stats-bfcl-v2 | BFCL v2 | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/bfcl-v2?top_n=500 | Berkeley Function Calling Leaderboard (BFCL) v2 is a comprehensive benchmark for evaluating large language models' function calling capabilities. It features 2,251 question-function-answer pairs with enterprise and OSS-contributed functions, addressing data contamination and bias through live, user-contributed scenario... | [
"reasoning",
"general",
"tool_calling"
] | [] | text | null | null | unknown | null | null | null | 1 | 5 | 5 | 0.773 | null |
llm-stats:bfcl-v3 | llm-stats-bfcl-v3 | BFCL-v3 | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/bfcl-v3?top_n=500 | Berkeley Function Calling Leaderboard v3 (BFCL-v3) is an advanced benchmark that evaluates large language models' function calling capabilities through multi-turn and multi-step interactions. It introduces extended conversational exchanges where models must retain contextual information across turns and execute multipl... | [
"reasoning",
"structured_output",
"finance",
"general",
"agents",
"tool_calling"
] | [] | text | null | null | unknown | null | null | null | 1 | 19 | 19 | 0.778 | null |
llm-stats:bfcl-v4 | llm-stats-bfcl-v4 | BFCL-V4 | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/bfcl-v4?top_n=500 | Berkeley Function Calling Leaderboard V4 (BFCL-V4) evaluates LLMs on their ability to accurately call functions and APIs, including simple, multiple, parallel, and nested function calls across diverse programming scenarios. | [
"agents",
"tool_calling"
] | [] | text | null | null | unknown | null | null | null | 1 | 15 | 15 | 0.75 | null |
llm-stats:bfcl-v3-multiturn | llm-stats-bfcl-v3-multiturn | BFCL_v3_MultiTurn | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/bfcl-v3-multiturn?top_n=500 | Berkeley Function Calling Leaderboard (BFCL) V3 MultiTurn benchmark that evaluates large language models' ability to handle multi-turn and multi-step function calling scenarios. The benchmark introduces complex interactions requiring models to manage sequential function calls, handle conversational context across multi... | [
"reasoning",
"general",
"tool_calling"
] | [] | text | null | null | unknown | null | null | null | 1 | 2 | 2 | 0.768 | null |
llm-stats:big-bench-audio | llm-stats-big-bench-audio | Big Bench Audio | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/big-bench-audio?top_n=500 | Big Bench Audio is an audio reasoning benchmark adapted from a subset of Big Bench Hard, with text questions converted to spoken audio. It evaluates the reasoning ability of speech-to-speech and audio language models on tasks delivered as audio input, with accuracy scored by an independent evaluation (Artificial Analys... | [
"reasoning",
"audio"
] | [] | audio | null | null | unknown | null | null | null | 1 | 1 | 1 | 0.87 | null |
llm-stats:big-finance-bench | llm-stats-big-finance-bench | Big Finance Bench | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/big-finance-bench?top_n=500 | Big Finance Bench evaluates models on complex financial-analysis tasks that require retrieving and reasoning over financial documents and performing multi-step quantitative work. | [
"reasoning",
"finance",
"agents"
] | [] | text | null | null | unknown | null | null | null | 1 | 3 | 3 | 0.53 | null |
llm-stats:big-bench | llm-stats-big-bench | BIG-Bench | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/big-bench?top_n=500 | Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark consisting of 204+ tasks designed to probe large language models and extrapolate their future capabilities. It covers diverse domains including linguistics, mathematics, common-sense reasoning, biology, physics, social bias, software developme... | [
"math",
"reasoning",
"language"
] | [] | text | null | null | unknown | null | null | null | 1 | 3 | 3 | 0.75 | null |
llm-stats:big-bench-extra-hard | llm-stats-big-bench-extra-hard | BIG-Bench Extra Hard | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/big-bench-extra-hard?top_n=500 | BIG-Bench Extra Hard (BBEH) is a challenging benchmark that replaces each task in BIG-Bench Hard with a novel task that probes similar reasoning capabilities but exhibits significantly increased difficulty. The benchmark contains 23 tasks testing diverse reasoning skills including many-hop reasoning, causal understandi... | [
"reasoning",
"language",
"general"
] | [] | text | null | null | unknown | null | null | null | 1 | 11 | 11 | 0.744 | null |
model-reports:bigbench_extra_hard | bigbench_extra_hard | BIG-Bench Extra Hard | model_reports | https://arxiv.org/abs/2502.19187 | Successor to BBH after saturation; per-task variance is high. | [
"reasoning"
] | [] | null | null | 2025-02-26T00:00:00 | unknown | https://arxiv.org/abs/2502.19187 | null | null | 1 | 1 | 1 | 74.4 | percent |
llm-stats:big-bench-hard | llm-stats-big-bench-hard | BIG-Bench Hard | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/big-bench-hard?top_n=500 | BIG-Bench Hard (BBH) is a subset of 23 challenging BIG-Bench tasks selected because prior language model evaluations did not outperform average human-rater performance. The benchmark contains 6,511 evaluation examples testing various forms of multi-step reasoning including arithmetic, logical reasoning (Boolean express... | [
"math",
"reasoning",
"language"
] | [] | text | null | null | unknown | null | https://github.com/suzgunmirac/BIG-Bench-Hard | null | 1 | 21 | 21 | 0.931 | null |
llm-stats:bigcodebench | llm-stats-bigcodebench | BigCodeBench | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/bigcodebench?top_n=500 | A benchmark that challenges LLMs to invoke multiple function calls as tools from 139 libraries and 7 domains for 1,140 fine-grained programming tasks. Evaluates code generation with diverse function calls and complex instructions, featuring two variants: Complete (code completion based on comprehensive docstrings) and ... | [
"reasoning",
"general"
] | [] | text | null | null | unknown | null | null | null | 1 | 2 | 2 | 0.454 | null |
opencompass:1253 | opencompass-1253-bigcodebench | BigCodeBench | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/BigCodeBench | BigCodeBench is a benchmark that challenges LLMs to invoke multiple function calls as tools from 139 libraries and 7 domains for 1,140 fine-grained tasks. BigCodeBench用于评估LLM的代码生成能力，包含1140个可以调用139个库和7个域的多个函数来完成的细粒度任务。 | [
"强推理",
"Strong Reasoning",
"代码",
"Code",
"大语言模型",
"LLM",
"逻辑推理",
"Reasoning",
"代码工程",
"开源收录",
"Open-Source",
"不支持",
"Unsupported"
] | [] | null | Monash University | 2024-06-22T00:00:00 | open | https://arxiv.org/abs/2406.15877 | https://github.com/bigcode-project/bigcodebench | https://huggingface.co/datasets/bigcode/bigcodebench | 1 | null | 0 | null | null |
llm-stats:bigcodebench-full | llm-stats-bigcodebench-full | BigCodeBench-Full | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/bigcodebench-full?top_n=500 | A comprehensive benchmark that evaluates large language models' ability to solve complex, practical programming tasks via code generation. Contains 1,140 fine-grained tasks across 7 domains using function calls from 139 libraries. Challenges LLMs to invoke multiple function calls as tools and handle complex instruction... | [
"reasoning",
"general"
] | [] | text | null | null | unknown | null | null | null | 1 | 1 | 1 | 0.496 | null |
llm-stats:bigcodebench-hard | llm-stats-bigcodebench-hard | BigCodeBench-Hard | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/bigcodebench-hard?top_n=500 | BigCodeBench-Hard is a subset of 148 challenging programming tasks from BigCodeBench, designed to evaluate large language models' ability to solve complex, real-world programming problems. These tasks require diverse function calls from multiple libraries across 7 domains including computation, networking, data analysi... | [
"reasoning",
"general"
] | [] | text | null | null | unknown | null | null | null | 1 | 1 | 1 | 0.27 | null |
opencompass:1680 | opencompass-1680-bigobench | BigOBench | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/BigOBench | BigO(Bench)是一个包含约 300 个需要用 Python 解决的代码问题的基准测试，以及 3,105 个编码问题和 1,190,250 个解决方案用于训练，以评估LLMs能否找到代码解决方案的时间-空间复杂度，或者生成符合时间-空间复杂度要求的代码解决方案。 BigO(Bench)是一个包含约 300 个需要用 Python 解决的代码问题的基准测试，以及 3,105 个编码问题和 1,190,250 个解决方案用于训练，以评估LLMs能否找到代码解决方案的时间-空间复杂度，或者生成符合时间-空间复杂度要求的代码解决方案。 | [
"代码",
"Code",
"大语言模型",
"LLM",
"代码工程",
"不支持",
"Unsupported"
] | [] | null | facebook | 2025-03-19T00:00:00 | restricted | https://arxiv.org/abs/2503.15242 | https://github.com/facebookresearch/bigobench | https://huggingface.co/datasets/facebook/BigOBench | 1 | null | 0 | null | null |
llm-stats:biolp-bench | llm-stats-biolp-bench | BioLP-Bench | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/biolp-bench?top_n=500 | BioLP-Bench is a model-graded evaluation measuring ability to find and correct mistakes in common biological laboratory protocols. It evaluates dual-use biological knowledge relevant to bioweapons development. | [
"safety",
"healthcare",
"biology"
] | [] | text | null | null | unknown | null | null | null | 1 | 1 | 1 | 0.37 | null |
llm-stats:biomysterybench | llm-stats-biomysterybench | BioMysteryBench | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/biomysterybench?top_n=500 | BioMysteryBench evaluates a model's ability to reason through challenging molecular biology problems, reporting performance on a hard subset and on the subset of problems solved by human experts. | [
"reasoning",
"science",
"biology"
] | [] | text | null | null | unknown | null | null | null | 1 | 3 | 3 | 0.901 | null |
model-reports:biomysterybench | biomysterybench | BioMysteryBench | model_reports | https://huggingface.co/datasets/Anthropic/BioMysteryBench-full | Reported in two splits ("hard" and "human solved") that differ by more than 35 points, so a bare score is unreadable without its split. Anthropic notes its own safety refusals depress this number, which means the score mixes capability with policy. First-party to Anthropic, which built it and publishes the dataset. The... | [
"biology"
] | [] | null | null | 2026-04-29T00:00:00 | unknown | null | null | null | 2 | 1 | 1 | 71.3 | percent |
llm-stats:bird-sql-(dev) | llm-stats-bird-sql-dev | Bird-SQL (dev) | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/bird-sql-%28dev%29?top_n=500 | BIRD (BIg Bench for LaRge-scale Database Grounded Text-to-SQLs) is a comprehensive text-to-SQL benchmark containing 12,751 question-SQL pairs across 95 databases (33.4 GB total) spanning 37+ professional domains. It evaluates large language models' ability to convert natural language to executable SQL queries in real-w... | [
"reasoning"
] | [] | text | null | null | unknown | null | null | null | 1 | 7 | 7 | 0.574 | null |
llm-stats:bixbench | llm-stats-bixbench | BixBench | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/bixbench?top_n=500 | BixBench is a benchmark for real-world bioinformatics and computational biology data analysis. It evaluates AI models on multi-step scientific workflows that require code execution, statistical reasoning, and biological domain knowledge to interpret experimental data. | [
"reasoning",
"science",
"agents"
] | [] | text | null | null | unknown | null | null | null | 1 | 1 | 1 | 0.805 | null |
llm-stats:blink | llm-stats-blink | BLINK | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/blink?top_n=500 | BLINK: Multimodal Large Language Models Can See but Not Perceive. A benchmark for multimodal language models focusing on core visual perception abilities. Reformats 14 classic computer vision tasks into 3,807 multiple-choice questions paired with single or multiple images and visual prompting. Tasks include relative de... | [
"multimodal",
"reasoning",
"spatial_reasoning",
"3d",
"vision"
] | [] | multimodal | null | null | unknown | null | null | null | 1 | 15 | 15 | 0.814 | null |
opencompass:1365 | opencompass-1365-blink | BLINK | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/BLINK | BLINK focuses on MLLMs' core visual perception abilities. It contains 3,807 multiple-choice questions spanning 14 classic computer vision tasks. BLINK用于评估多模态大模型的视觉感知能力，包含来自14个经典计算机视觉任务的3807道多项选择题。 | [
"多模态",
"Multimodal",
"理解",
"Understanding",
"多模态模型",
"VLM",
"图像理解",
"Image Understanding",
"不支持",
"Unsupported"
] | [] | multimodal | University of Pensylvania | 2024-04-28T00:00:00 | open | https://arxiv.org/abs/2404.12390 | https://github.com/zeyofu/BLINK_Benchmark | https://huggingface.co/datasets/BLINK-Benchmark/BLINK | 1 | null | 0 | null | null |
llm-stats:blueprint-bench-2 | llm-stats-blueprint-bench-2 | Blueprint-Bench 2 | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/blueprint-bench-2?top_n=500 | Blueprint-Bench 2 is an agentic spatial reasoning benchmark that evaluates a model's ability to understand, plan, and reason over architectural blueprints and other structured spatial documents. Scores are reported as a normalized score. | [
"multimodal",
"reasoning",
"agents"
] | [] | multimodal | null | null | unknown | null | null | null | 1 | 2 | 2 | 0.386 | null |
model-reports:blueprint_bench_2 | blueprint_bench_2 | Blueprint-Bench 2 | model_reports | https://andonlabs.com/evals/blueprint-bench-2 | Spatial-reasoning set: 50 apartments, ~20 photos each, scored by a connectivity-graph grader, with a public leaderboard on the Andon Labs eval page (run in-house; dataset not openly downloadable). Builds on the original Blueprint-Bench paper (arXiv:2509.25229), a distinct instrument released 2025-09-24 whose scores mus... | [
"spatial_reasoning"
] | [] | null | null | 2026-05-04T00:00:00 | unknown | null | null | null | 1 | null | 0 | null | null |
llm-stats:boolq | llm-stats-boolq | BoolQ | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/boolq?top_n=500 | BoolQ is a reading comprehension dataset for yes/no questions containing 15,942 naturally occurring examples. Each example consists of a question, passage, and boolean answer, where questions are generated in unprompted and unconstrained settings. The dataset challenges models with complex, non-factoid information requ... | [
"reasoning",
"language"
] | [] | text | null | null | unknown | null | null | null | 1 | 10 | 10 | 0.8804 | null |
opencompass:510 | opencompass-510-boolq | BoolQ | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/BoolQ | BoolQ is a question answering dataset for yes/no questions containing 15942 examples. These questions are naturally occurring ---they are generated in unprompted and unconstrained settings. Each example is a triplet of (question, passage, answer), with the title of the page as optional additional context. BoolQ是一个包含159... | [
"知识",
"Knowledge",
"大语言模型",
"LLM",
"知识储备",
"开源收录",
"Open-Source",
"不支持",
"Unsupported"
] | [] | null | null | 2019-05-24T00:00:00 | unknown | https://arxiv.org/abs/1905.10044 | https://github.com/google-research-datasets/boolean-questions | null | 1 | null | 0 | null | null |
opencompass:1571 | opencompass-1571-bright | BRIGHT | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/BRIGHT | BRIGHT is the first text retrieval benchmark that requires intensive reasoning to retrieve relevant documents. BRIGHT 是第一个需要大量推理来检索相关文档的文本检索基准。 | [
"强推理",
"Strong Reasoning",
"大语言模型",
"LLM",
"检索能力",
"Retrieval",
"逻辑推理",
"Reasoning",
"不支持",
"Unsupported"
] | [] | null | The University of Hong Kong, etc. | 2024-10-24T00:00:00 | unknown | https://arxiv.org/abs/2407.12883 | https://github.com/xlang-ai/BRIGHT | https://huggingface.co/datasets/xlangai/BRIGHT | 1 | null | 0 | null | null |
model-reports:brokenarxiv | brokenarxiv | BrokenArXiv | model_reports | https://matharena.ai/brokenarxiv | arXiv proofs with planted errors; tests whether a model catches a broken argument instead of reproducing it. | [
"math"
] | [] | null | null | null | unknown | null | null | null | 1 | 1 | 1 | 54.6 | percent |
llm-stats:browsecomp | llm-stats-browsecomp | BrowseComp | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/browsecomp?top_n=500 | BrowseComp is a benchmark comprising 1,266 questions that challenge AI agents to persistently navigate the internet in search of hard-to-find, entangled information. The benchmark measures agents' ability to exercise persistence in information gathering, demonstrate creativity in web navigation, and find concise, verif... | [
"reasoning",
"search",
"agents"
] | [] | text | OpenAI | 2025-04-10T00:00:00 | restricted | https://arxiv.org/abs/2504.12516 | https://github.com/openai/simple-evals | null | 1 | 62 | 62 | 0.912 | null |
model-reports:browsecomp | browsecomp | BrowseComp | model_reports | https://openai.com/index/browsecomp/ | Live web. The result depends on what the internet contained on the day of the run. | [
"agent"
] | [] | null | null | 2025-04-10T00:00:00 | unknown | null | null | null | 12 | 7 | 9 | 91.2 | percent |
llm-stats:browsecomp-long-128k | llm-stats-browsecomp-long-128k | BrowseComp Long Context 128k | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/browsecomp-long-128k?top_n=500 | A challenging benchmark for evaluating web browsing agents' ability to persistently navigate the internet and find hard-to-locate, entangled information. Comprises 1,266 questions requiring strategic reasoning, creative search, and interpretation of retrieved content, with short and easily verifiable answers. | [
"reasoning",
"search"
] | [] | text | null | null | unknown | null | null | null | 1 | 5 | 5 | 0.92 | null |
llm-stats:browsecomp-long-256k | llm-stats-browsecomp-long-256k | BrowseComp Long Context 256k | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/browsecomp-long-256k?top_n=500 | BrowseComp is a benchmark for measuring the ability of agents to browse the web, comprising 1,266 questions that require persistently navigating the internet in search of hard-to-find, entangled information. Despite the difficulty of the questions, BrowseComp is simple and easy-to-use, as predicted answers are short an... | [
"reasoning",
"search"
] | [] | text | null | null | unknown | null | null | null | 1 | 2 | 2 | 0.898 | null |
llm-stats:browsecomp-vl | llm-stats-browsecomp-vl | BrowseComp-VL | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/browsecomp-vl?top_n=500 | BrowseComp-VL is the vision-language variant of BrowseComp, evaluating multimodal models on web browsing comprehension tasks that require processing visual web page content alongside text. | [
"multimodal",
"search",
"agents",
"vision"
] | [] | multimodal | null | null | unknown | null | null | null | 1 | 1 | 1 | 0.519 | null |
llm-stats:browsecomp-zh | llm-stats-browsecomp-zh | BrowseComp-zh | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/browsecomp-zh?top_n=500 | A high-difficulty benchmark purpose-built to comprehensively evaluate LLM agents on the Chinese web, consisting of 289 multi-hop questions spanning 11 diverse domains including Film & TV, Technology, Medicine, and History. Questions are reverse-engineered from short, objective, and easily verifiable answers, requiring ... | [
"reasoning",
"search"
] | [] | text | null | null | unknown | null | null | null | 1 | 13 | 13 | 0.703 | null |
model-reports:browsecomp_zh | browsecomp_zh | BrowseComp-ZH | model_reports | https://arxiv.org/abs/2504.19314 | Chinese-language live web. Same day-to-day web drift as BrowseComp, over a different index. | [
"agent"
] | [] | null | null | 2025-04-27T00:00:00 | unknown | https://arxiv.org/abs/2504.19314 | null | null | 2 | 2 | 2 | 72.7 | percent |
opencompass:1146 | opencompass-1146-bust | BUST | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/BUST | BUST is a comprehensive benchmark for evaluating synthetic text detectors, focusing on their effectiveness against outputs from various Large Language Models (LLMs). BUST 是一个综合基准，旨在评估合成文本检测器，BUST 使用多种指标来评估检测器，包括语言特征、可读性和作者态度。 | [
"理解",
"Understanding",
"NAACL 2024",
"大语言模型",
"LLM",
"语言理解",
"Comprehension",
"不支持",
"Unsupported"
] | [] | null | Dalle Molle Institute for Artificial Intelligence Research (IDSIA) | 2024-06-16T00:00:00 | unknown | null | https://github.com/IDSIA-NLP/BUST | null | 1 | null | 0 | null | null |
opencompass:1983 | opencompass-1983-bytemorph | ByteMorph | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/ByteMorph | ByteMorph is a benchmark for instruction-guided image editing, focusing on evaluating models’ capabilities in handling non-rigid motions such as camera viewpoint changes, object deformations, human articulations, and complex interactions. ByteMorph 是一个面向指令驱动图像编辑的基准，专注于评估模型在处理非刚性运动（如相机视角变化、物体变形、人类动作和复杂交互）方面的能力。 该基准包括超过 ... | [
"创作",
"Creation",
"指令跟随",
"Instruct",
"多模态模型",
"VLM",
"视觉生成",
"Visual Generation",
"指令遵循",
"Instruction Following",
"不支持",
"Unsupported"
] | [] | null | ByteDance Seed , University of Southern California , University of Tokyo , etc. | 2025-06-03T00:00:00 | open | https://arxiv.org/abs/2506.03107 | https://github.com/ByteDance-Seed/BM-code | https://huggingface.co/datasets/ByteDance-Seed/BM-6M | 1 | null | 0 | null | null |
llm-stats:c-eval | llm-stats-c-eval | C-Eval | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/c-eval?top_n=500 | C-Eval is a comprehensive Chinese evaluation suite designed to assess advanced knowledge and reasoning abilities of foundation models in a Chinese context. It comprises 13,948 multiple-choice questions across 52 diverse disciplines spanning humanities, science, and engineering, with four difficulty levels: middle schoo... | [
"reasoning",
"general"
] | [] | text | null | null | unknown | null | null | null | 1 | 18 | 18 | 0.933 | null |
opencompass:496 | opencompass-496-c-eval | C-Eval | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/C-Eval | C-Eval is a comprehensive Chinese evaluation suite for foundation models. It consists of 13948 multi-choice questions spanning 52 diverse disciplines and four difficulty levels. C-Eval 是一个全面的中文基础模型评估套件。它包含了13948个多项选择题，涵盖了52个不同的学科和四个难度级别。 | [
"学科",
"Examination",
"大语言模型",
"LLM",
"知识储备",
"Knowledge",
"开源收录",
"Open-Source",
"支持",
"Supported"
] | [
"English",
"Chinese"
] | null | null | 2023-05-15T00:00:00 | unknown | https://arxiv.org/abs/2305.08322 | https://github.com/SJTU-LIT/ceval | null | 1 | null | 0 | null | null |
opencompass:1780 | opencompass-1780-c-faith | C-FAITH | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/C-FAITH | C-FAITH, a Chinese QA hallucination benchmark created from 1,399 knowledge documents obtained from web scraping, totaling 60,702 entries. C-FAITH，这是一个中国 QA 幻觉基准，由从网络抓取中获得的 1,399 份知识文档创建，总共 60,702 个条目。 | [
"语言",
"Language",
"大语言模型",
"LLM",
"事实可靠性",
"Factual Reliability",
"不支持",
"Unsupported"
] | [
"Chinese"
] | null | PKU | 2025-04-14T00:00:00 | unknown | https://arxiv.org/abs/2504.10167 | https://github.com/pkulcwmzx/C-FAITH | null | 1 | null | 0 | null | null |
opencompass:514 | opencompass-514-c3 | C3 | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/C3 | A free-form multiple-Choice Chinese machine reading Comprehension dataset (C3), containing 13,369 documents (dialogues or more formally written mixed-genre texts) and their associated 19,577 multiple-choice free-form questions collected from Chinese-as-a-second-language examinations 一个自由形式的多项选择中文机器阅读理解数据集（C3），包含13369篇文... | [
"理解",
"Understanding",
"大语言模型",
"LLM",
"语言理解",
"Comprehension",
"开源收录",
"Open-Source",
"不支持",
"Unsupported"
] | [
"Chinese"
] | null | null | 2019-04-21T00:00:00 | unknown | https://arxiv.org/abs/1904.09679v3 | https://github.com/nlpdata/c3 | null | 1 | null | 0 | null | null |
opencompass:1052 | opencompass-1052-calm | CaLM | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/CaLM | CaLM is the first comprehensive benchmark for evaluating the causal reasoning capabilities of language models. The CaLM framework establishes a foundational taxonomy consisting of four modules: causal target, adaptation, metric, and error. CaLM是上海人工智能实验室联合同济大学、上海交通大学、北京大学及商汤科技发布首个大模型因果推理开放评测体系。首次从因果推理角度提出评估框架，为AI研究者打造可... | [
"推理",
"Reasoning",
"大语言模型",
"LLM",
"逻辑推理",
"开源收录",
"Open-Source",
"不支持",
"Unsupported"
] | [
"English",
"Chinese"
] | null | Shanghai AI Laboratory | 2024-05-01T00:00:00 | unknown | https://arxiv.org/abs/2405.00622 | https://github.com/OpenCausaLab/CaLM | null | 1 | null | 0 | null | null |
llm-stats:capture-the-flag-challenges | llm-stats-capture-the-flag-challenges | Capture-the-Flag Challenges (Internal) | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/capture-the-flag-challenges?top_n=500 | Capture-the-Flag Challenges is OpenAI's internal expansion of competitive, professional-level cybersecurity capture-the-flag tasks used to evaluate vulnerability identification and exploitation capability under the Preparedness Framework. | [
"safety",
"agents",
"code"
] | [] | text | null | null | unknown | null | null | null | 1 | 3 | 3 | 0.967 | null |
opencompass:1975 | opencompass-1975-causalvqa | CausalVQA | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/CausalVQA | CausalVQA tests causal reasoning in videos across five question types, and state-of-the-art multimodal models still trail human performance. CausalVQA 是面向视频问答的因果推理基准，涵盖反事实、假设、预判、规划、描述五类问题，强调真实物理场景。 | [
"多模态",
"Multimodal",
"VQA",
"多模态模型",
"VLM",
"跨模态推理",
"Cross-modal Reasoning",
"不支持",
"Unsupported"
] | [] | multimodal | FAIR at Meta | 2025-06-11T00:00:00 | unknown | https://arxiv.org/abs/2506.09943 | https://github.com/facebookresearch/CausalVQA | null | 1 | null | 0 | null | null |
llm-stats:cbnsl | llm-stats-cbnsl | CBNSL | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/cbnsl?top_n=500 | Curriculum Learning of Bayesian Network Structures (CBNSL) benchmark for evaluating algorithms that learn Bayesian network structures from data using curriculum learning techniques. The benchmark uses networks from the bnlearn repository and evaluates structure learning performance using BDeu scoring metrics. | [
"math",
"reasoning"
] | [] | text | null | null | unknown | null | null | null | 1 | 1 | 1 | 0.956 | null |
llm-stats:cc-bench-v2-backend | llm-stats-cc-bench-v2-backend | CC-Bench-V2 Backend | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/cc-bench-v2-backend?top_n=500 | CC-Bench-V2 Backend evaluates coding agents on backend development tasks, measuring practical engineering ability to implement server-side logic, APIs, and system components. | [
"code"
] | [] | text | null | null | unknown | null | null | null | 1 | 1 | 1 | 0.228 | null |
llm-stats:cc-bench-v2-frontend | llm-stats-cc-bench-v2-frontend | CC-Bench-V2 Frontend | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/cc-bench-v2-frontend?top_n=500 | CC-Bench-V2 Frontend evaluates coding agents on frontend development tasks, measuring ability to build UI components, handle styling, and implement client-side logic. | [
"code"
] | [] | text | null | null | unknown | null | null | null | 1 | 1 | 1 | 0.684 | null |
llm-stats:cc-bench-v2-repo | llm-stats-cc-bench-v2-repo | CC-Bench-V2 Repo Exploration | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/cc-bench-v2-repo?top_n=500 | CC-Bench-V2 Repo Exploration evaluates coding agents on repository-level understanding and navigation, measuring ability to explore, comprehend, and work across entire codebases. | [
"agents",
"code"
] | [] | text | null | null | unknown | null | null | null | 1 | 1 | 1 | 0.722 | null |
llm-stats:cc-ocr | llm-stats-cc-ocr | CC-OCR | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/cc-ocr?top_n=500 | A comprehensive OCR benchmark for evaluating Large Multimodal Models (LMMs) in literacy. Comprises four OCR-centric tracks: multi-scene text reading, multilingual text reading, document parsing, and key information extraction. Contains 39 subsets with 7,058 fully annotated images, 41% sourced from real applications. Te... | [
"multimodal",
"structured_output",
"text-to-image",
"vision"
] | [] | multimodal | null | null | unknown | null | null | null | 1 | 18 | 18 | 0.834 | null |
opencompass:1552 | opencompass-1552-ceb | CEB | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/CEB | CEB evaluates LLM bias compositionally, featuring 11k samples characterized across bias types, social groups, and tasks. CEB是一个用于大型语言模型偏差的组成评估基准，引入了包含 11,004 个样本的组成评估基准，从偏差类型、社会群体和任务三个维度描述每个数据集。 | [
"其他",
"Other",
"大语言模型",
"LLM",
"安全对齐",
"Safety Alignment",
"不支持",
"Unsupported"
] | [] | null | University of Virginia, Arizona State University, etc. | 2024-07-03T00:00:00 | restricted | https://arxiv.org/abs/2407.02408 | https://github.com/SongW-SW/CEB | https://huggingface.co/datasets/Song-SW/CEB | 1 | null | 0 | null | null |
llm-stats:cfeval | llm-stats-cfeval | CFEval | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/cfeval?top_n=500 | CFEval benchmark for evaluating code generation and problem-solving capabilities | [
"code"
] | [] | text | null | null | unknown | null | null | null | 1 | 2 | 2 | 2,134 | null |
opencompass:1907 | opencompass-1907-cfinbench | CFinBench | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/CFinBench | We present CFinBench: a meticulously crafted, the most comprehensive evaluation benchmark to date, for assessing the financial knowledge of LLMs under Chinese context. 为了更加全面地探究大语言模型在中文财经领域的能力，本工作提出了目前为止量级最大的中文财经评测基准（CFinBench）。该数据集共包含99,100个评测样本，并包含单选题、多选题和判断题在内的三种题型。该工作对当前主流的大模型从四个维度进行了详细评测：财经学科基础、财经资格认证、财经从业实践、财经法律法... | [
"知识",
"Knowledge",
"NAACL 2025",
"金融",
"大语言模型",
"LLM",
"知识储备",
"不支持",
"Unsupported"
] | [
"Chinese"
] | null | 华为, 新加坡南洋理工 | 2024-10-01T00:00:00 | unknown | null | null | null | 1 | null | 0 | null | null |
opencompass:1079 | opencompass-1079-cflue | CFLUE | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/CFLUE | CFLUE is the Chinese Financial Language Understanding Evaluation benchmark, designed to assess the capability of LLMs across various dimensions. CFLUE 是中国金融语言理解评估基准，旨在评估大型语言模型（LLMs）在各个维度上的能力。具体而言，CFLUE 提供了针对知识评估和应用评估量身定制的数据集。在知识评估方面，它包含超过 38,000 道选择题及相关的解决方案解释。 | [
"知识",
"Knowledge",
"ACL 2024",
"大语言模型",
"LLM",
"知识储备",
"不支持",
"Unsupported"
] | [] | null | Alibaba | 2024-08-11T00:00:00 | unknown | null | https://github.com/aliyun/cflue | null | 1 | null | 0 | null | null |
opencompass:1513 | opencompass-1513-cg-bench | CG-Bench | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/CG-Bench | CG-Bench is meant for evaluating MLLMs' long video understanding, including 12,129 QA pairs from 1219 videos in 3 major question types: perception, reasoning, and hallucination. CG-Bench用于评估多模态大模型的长视频理解能力，基于1219个视频设计了12129个涵盖感知、推理和幻觉三种问题类型的QA对。 | [
"多模态",
"Multimodal",
"理解",
"Understanding",
"多模态模型",
"VLM",
"视频理解",
"Video Understanding",
"不支持",
"Unsupported"
] | [] | multimodal | Nanjing University | 2024-12-16T00:00:00 | restricted | https://arxiv.org/abs/2412.12075v1 | https://github.com/CG-Bench/CG-Bench | https://huggingface.co/datasets/CG-Bench/CG-Bench | 1 | null | 0 | null | null |
llm-stats:charadessta | llm-stats-charadessta | CharadesSTA | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/charadessta?top_n=500 | Charades-STA is a benchmark dataset for temporal activity localization via language queries, extending the Charades dataset with sentence temporal annotations. It contains 12,408 training and 3,720 testing segment-sentence pairs from videos with natural language descriptions and precise temporal boundaries for localizi... | [
"multimodal",
"language",
"video",
"vision"
] | [] | multimodal | null | null | unknown | null | null | null | 1 | 12 | 12 | 0.648 | null |
opencompass:1141 | opencompass-1141-charm | CHARM | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/CHARM | CHARM is the first benchmark for comprehensively and in-depth evaluating the commonsense reasoning ability of large language models (LLMs) in Chinese, which covers both globally known and Chinese-specific commonsense. CHARM 是首个全面深入评估大语言模型（LLMs）在中文中的常识推理能力的基准，涵盖了全球通用的常识和特定于中国的常识。 | [
"推理",
"Reasoning",
"ACL 2024",
"大语言模型",
"LLM",
"逻辑推理",
"不支持",
"Unsupported"
] | [
"Chinese"
] | null | Shanghai AI Laboratory | 2024-08-11T00:00:00 | unknown | https://arxiv.org/abs/1809.05053 | https://github.com/opendatalab/CHARM | null | 1 | null | 0 | null | null |
llm-stats:chartmuseum | llm-stats-chartmuseum | ChartMuseum | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/chartmuseum?top_n=500 | ChartMuseum is a chart question-answering benchmark of 1,162 expert-annotated questions over real-world chart images drawn from 184 sources, including academic figures, infographics, and unconventional chart designs. It specifically targets questions that require visual reasoning, such as comparing unlabeled visual ele... | [
"multimodal",
"reasoning",
"vision"
] | [] | multimodal | null | null | unknown | null | null | null | 1 | 1 | 1 | 0.867 | null |
model-reports:chartography | chartography | Chartography | model_reports | https://github.com/Chartography/Chartography | Chart comprehension benchmark with tool access. Scores depend on context length and tool configuration. | [
"vision"
] | [] | null | null | 2025-06-01T00:00:00 | unknown | null | https://github.com/Chartography/Chartography | null | 1 | 1 | 1 | 78 | percent |
llm-stats:chartqa | llm-stats-chartqa | ChartQA | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/chartqa?top_n=500 | ChartQA is a large-scale benchmark comprising 9.6K human-written questions and 23.1K questions generated from human-written chart summaries, designed to evaluate models' abilities in visual and logical reasoning over charts. | [
"multimodal",
"reasoning",
"vision"
] | [] | multimodal | null | null | unknown | https://aclanthology.org/2022.findings-acl.177.pdf | https://github.com/vis-nlp/ChartQA | null | 1 | 26 | 26 | 0.908 | null |
model-reports:chartqa | chartqa | ChartQA | model_reports | https://github.com/vis-nlp/ChartQA | Largely saturated; relaxed-accuracy tolerance affects the reported figure. | [
"multimodal"
] | [] | null | null | 2022-03-19T00:00:00 | unknown | null | https://github.com/vis-nlp/ChartQA | null | 2 | null | 0 | null | null |
llm-stats:chartqapro | llm-stats-chartqapro | ChartQAPro | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/chartqapro?top_n=500 | ChartQAPro is a challenging benchmark for question answering over diverse, real-world charts and infographics. | [
"multimodal",
"reasoning",
"vision"
] | [] | multimodal | null | null | unknown | null | null | null | 1 | 2 | 2 | 0.709 | null |
model-reports:charxiv_reasoning | charxiv_reasoning | CharXiv Reasoning | model_reports | https://github.com/CharXiv/CharXiv | Chart reasoning benchmark with tool access. Scores depend on context length and tool configuration. | [
"vision"
] | [] | null | null | 2025-06-01T00:00:00 | unknown | null | https://github.com/CharXiv/CharXiv | null | 1 | 1 | 1 | 89.4 | percent |
llm-stats:charxiv-d | llm-stats-charxiv-d | CharXiv-D | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/charxiv-d?top_n=500 | CharXiv-D is the descriptive questions subset of the CharXiv benchmark, designed to assess multimodal large language models' ability to extract basic information from scientific charts. It contains descriptive questions covering information extraction, enumeration, pattern recognition, and counting across 2,323 diverse... | [
"multimodal",
"reasoning",
"structured_output",
"vision"
] | [] | multimodal | null | null | unknown | null | null | null | 1 | 17 | 17 | 0.955 | null |
llm-stats:charxiv-r | llm-stats-charxiv-r | CharXiv-R | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/charxiv-r?top_n=500 | CharXiv-R is the reasoning component of the CharXiv benchmark, focusing on complex reasoning questions that require synthesizing information across visual chart elements. It evaluates multimodal large language models on their ability to understand and reason about scientific charts from arXiv papers through various rea... | [
"multimodal",
"reasoning",
"vision"
] | [] | multimodal | null | null | unknown | null | https://github.com/princeton-nlp/CharXiv | null | 1 | 51 | 51 | 0.932 | null |
opencompass:1534 | opencompass-1534-chase-code | CHASE-Code | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/CHASE-Code | CHASE is a unified framework to synthetically generate challenging problems using LLMs without human involvement CHASE是一个无需人工参与的统一框架，用于合成生成具有挑战性的问题 | [
"代码",
"Code",
"数学",
"Math",
"大语言模型",
"LLM",
"代码工程",
"数理能力",
"不支持",
"Unsupported"
] | [] | null | Mila and McGill University | 2025-02-14T00:00:00 | open | https://arxiv.org/pdf/2502.14678 | https://github.com/McGill-NLP/CHASE | https://huggingface.co/datasets/McGill-NLP/CHASE-Code | 1 | null | 0 | null | null |
model-reports:chatbot_arena | chatbot_arena | Chatbot Arena | model_reports | https://lmarena.ai/ | Elo from real user traffic; sampling and prompt distribution are outside any vendor's control. | [
"human_preference"
] | [] | null | null | 2023-05-03T00:00:00 | unknown | null | null | null | 1 | null | 0 | null | null |
opencompass:692 | opencompass-692-chembench | ChemBench | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/ChemBench | ChemBench is a large-scale chemistry competency evaluation benchmark for language models, which includes nine chemistry core tasks and 4100 high-quality single-choice questions and answers. ChemBench是一个包含了九项化学核心任务，4100个高质量单选问答的大语言模型化学能力评测基准. | [
"知识",
"Knowledge",
"科学智能",
"AI for Science",
"知识储备",
"科学推理",
"Scientific Reasoning",
"合作共建",
"Co-Built",
"不支持",
"Unsupported"
] | [] | null | null | 2024-02-15T00:00:00 | unknown | https://arxiv.org/abs/2402.06852 | null | https://huggingface.co/datasets/AI4Chem/ChemBench4K | 1 | null | 0 | null | null |
llm-stats:chexpert-cxr | llm-stats-chexpert-cxr | CheXpert CXR | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/chexpert-cxr?top_n=500 | CheXpert is a large dataset of 224,316 chest radiographs from 65,240 patients for automated chest X-ray interpretation. The dataset includes uncertainty labels for 14 medical observations extracted from radiology reports. It serves as a benchmark for developing and evaluating automated chest radiograph interpretation m... | [
"healthcare",
"vision"
] | [] | image | null | null | unknown | null | null | null | 1 | 1 | 1 | 0.481 | null |
opencompass:505 | opencompass-505-chid | CHID | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/CHID | CHID is a chinese idiom reading comprehension task, which requires to select the correct idiom to fill in the blank according to the context, with 10 candidate idioms. CHID是一个中文成语阅读理解任务，要求根据上下文选择正确的成语填空，共有10个候选成语。 | [
"语言",
"Language",
"大语言模型",
"LLM",
"语言理解",
"Comprehension",
"开源收录",
"Open-Source",
"不支持",
"Unsupported"
] | [
"Chinese"
] | null | null | 2019-06-04T00:00:00 | unknown | https://arxiv.org/abs/1906.01265 | https://github.com/chujiezheng/ChID-Dataset | null | 1 | null | 0 | null | null |
opencompass:1278 | opencompass-1278-chronomagic-bench | ChronoMagic-Bench | opencompass_hub | https://hub.opencompass.org.cn/dataset-detail/ChronoMagic-Bench | ChronoMagic-Bench can evaluate the temporal and metamorphic capabilities of the T2V (text-to-video) models in time-lapse video generation, introducing 1,649 prompts and real-world videos as references. ChronoMagic-Bench用来评估 T2V （文本到视频 ）模型在延时视频生成中的时间和变形能力，引入了1649个提示和真实世界的视频作为参考。 | [
"创作",
"Creation",
"NeurIPS 2024",
"多模态模型",
"VLM",
"视觉生成",
"Visual Generation",
"不支持",
"Unsupported"
] | [] | null | Peking University | 2024-06-26T00:00:00 | open | https://arxiv.org/abs/2406.18522 | https://github.com/PKU-YuanGroup/ChronoMagic-Bench | https://huggingface.co/datasets/BestWishYsh/ChronoMagic | 1 | null | 0 | null | null |
llm-stats:ci-memories-coverage | llm-stats-ci-memories-coverage | CI Memories Coverage | llm_stats | https://api.zeroeval.com/leaderboard/benchmarks/ci-memories-coverage?top_n=500 | CI Memories measures privacy behavior in memory-enabled agents using contextual-integrity scenarios. This metric reports evaluation coverage. | [
"memory",
"privacy",
"safety",
"agents"
] | [] | text | null | null | unknown | null | null | null | 1 | 1 | 1 | 0.648 | null |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.