LetTokenBurn
Sign in

Directory

Benchmarks

6,820 benchmarks indexed, 718 runnable through Harbor. Snapshot generated 2026-10-13. Each card shows a data-quality badge; verified, runnable entries are listed first.

  • adv glue

    Harbor adapter

    adv glue — AI evaluation benchmark listed by opencompass.

    Verified100%Safety & AlignmentKnowledge & Reasoning
    Samples
    121
    Released
    2021
    Scores
    0

  • AgroGround

    Harbor adapter

    A benchmark for grounded agricultural recognition measuring joint identity and localization correctness plus abstention on 1,480 human-verified images.

    Verified100%MultimodalKnowledge & Reasoning
    Samples
    1,480
    Released
    2026
    Scores
    5

  • ANLI

    Harbor adapter

    Adversarial NLI (ANLI) is a dataset collected via an iterative, adversarial human-and-model-in-the-loop procedure.

    Verified100%Knowledge & Reasoning
    Samples
    169,265
    Released
    2019
    Scores
    12

  • ArabicMMLU

    Harbor adapter

    The focus of language model evaluation has transitioned towards reasoning and knowledge intensive tasks, driven by advancements in pretraining large models.

    Verified100%Knowledge & Reasoning
    Samples
    74
    Released
    2024
    Scores
    0

  • ARC-AGI-2

    Harbor native

    ARC-AGI-2: 更难的抽象推理

    Verified100%Abstract Reasoning
    Samples
    167
    Released
    2025
    Scores
    118

  • ARC-Eval

    Harbor adapter

    ARC-Eval is a dataset of 7,787 multiple-choice science questions designed to evaluate reasoning capabilities in AI systems.

    Verified100%Knowledge & Reasoning
    Samples
    7,787
    Released
    2018
    Scores
    0

  • ASDiv

    Harbor adapter

    ASDiv (Academia Sinica Diverse MWP Dataset) is a diverse (in terms of both language patterns and problem types) English math word problem (MWP) corpus for evaluating the capability of various MWP solvers.

    Verified100%Math
    Samples
    2,305
    Released
    2021
    Scores
    0

  • Babilong

    Harbor adapter

    In recent years, the input context sizes of large language models (LLMs) have increased dramatically.

    Verified100%Long Context
    Samples
    1,000
    Released
    2024
    Scores
    0

  • BBH

    Harbor adapter

    BIG-Bench Hard evaluates advanced reasoning capabilities across 23 challenging tasks using over 6,500 English samples.

    Verified100%Knowledge & Reasoning
    Samples
    6,511
    Released
    2022
    Scores
    59

  • BBQ

    Harbor adapter

    BBQ: 社会偏见检测

    Verified100%Safety & Alignment
    Samples
    58,492
    Released
    2021
    Scores
    0

  • BigBench

    Harbor adapter

    The Beyond the Imitation Game Benchmark (BIG-bench) is a collaborative benchmark intended to probe large language models and extrapolate their future capabilities.

    Verified100%Knowledge & Reasoning
    Samples
    4,000
    Released
    2022
    Scores
    7

  • Bongard-OpenWorld

    Harbor adapter

    Bongard-OpenWorld is a new benchmark for evaluating real-world few-shot reasoning for machine vision.

    Verified100%MultimodalAbstract Reasoning
    Samples
    200
    Released
    2024
    Scores
    0

  • BoolQ

    Harbor adapter

    Reading comprehension dataset that queries for complex, non-factoid information, and require difficult entailment-like inference to solve.

    Verified100%Knowledge & Reasoning
    Samples
    3,270
    Released
    2019
    Scores
    162

  • BrowseComp

    Harbor adapter

    BrowseComp: 网页搜索研究任务

    Verified100%Agentic / Tool Use
    Samples
    1,266
    Released
    2025
    Scores
    6

  • C-Eval

    Harbor adapter

    C-Eval: 中文综合评估

    Verified100%Multilingual
    Samples
    13,948
    Released
    2023
    Scores
    0

  • Case HOLD

    Harbor adapter

    CaseHOLD (Case Holdings On Legal Decisions) is a multiple choice question answering scenario where the task is to identify the relevant holding of a cited case [(Zheng et al, 2021)](https://arxiv.org/pdf/2104.08671.pdf).

    Verified100%Knowledge & Reasoning
    Samples
    5,314
    Released
    2021
    Scores
    0

  • ChartQA

    Harbor adapter

    ChartQA: 图表理解问答

    Verified100%MultimodalMath
    Samples
    2,500
    Released
    2022
    Scores
    2

  • CodeContests

    Harbor native

    CodeContests: DeepMind编程竞赛

    Verified100%CodingMultilingual
    Samples
    9,644
    Released
    2022
    Scores
    1

  • commonsense_qa

    Harbor adapter

    CommonsenseQA is a multiple-choice question answering dataset that requires different types of commonsense knowledge to predict the correct answers.

    Verified100%Knowledge & Reasoning
    Samples
    1,221
    Released
    2019
    Scores
    5

  • DROP

    Harbor adapter

    DROP is a QA dataset which tests comprehensive understanding of paragraphs.

    Verified100%Knowledge & ReasoningMath
    Samples
    9,535
    Released
    2019
    Scores
    0

  • FrontierScience

    Harbor adapter

    FrontierScience evaluates AI models on expert-level scientific tasks derived from Olympiad problems and research literature to assess advanced reasoning capabilities.

    Verified100%Knowledge & Reasoning
    Samples
    160
    Released
    2026
    Scores
    0

  • GAIA

    Harbor native

    GAIA: 多步工具使用+推理

    Verified100%Agentic / Tool UseMultimodal
    Samples
    466
    Released
    2023
    Scores
    0

  • GPQA Diamond

    Harbor native

    GPQA Diamond is a graduate-level multiple-choice benchmark containing 198 questions in biology, physics, and chemistry designed to test models without reliance on search engines.

    Verified100%Knowledge & Reasoning
    Samples
    198
    Released
    2023
    Scores
    244