评测目录
查阅 1219 个已有官方记录的测试版本、指标与历史成绩。
常规展示测试
其他
- $OneMillion-Bench · Expert-score metric; task version and scale not stated
- 1H-VideoQA · Test subset
- τ²-bench · Anthropic corrected Airline variant; task and grading corrections submitted to evaluation authors
- τ²-Bench · Domain/version/metric unspecified in GLM source
- τ²-Bench · GLM-4.7模型卡;unknown
- τ²-Bench · GLM-5固定v2报告/模型卡;版本未注明
- τ²-bench · Retail section
- τ²-bench · Telecom section
- τ²-Bench · Weighted aggregate; release/version unspecified
- τ³-Bench · GLM-5.1模型卡;unknown
A
- AA-3D · Version and leaderboard snapshot time unspecified in release article
- AA-AnalystAgent · Version unspecified; all five independent attempts correct
- AA-Briefcase · Not stated in selected source
- AA-Briefcase · v1.1
- AA-LCR · Component of the Artificial Analysis Intelligence Index; version unspecified
- AA-LCR · 版本/单位未注明
- AA-Omniscience · Anthropic factuality evaluation on 42 economically relevant topics across six domains; Claude Sonnet 4.5 grader; dataset revision not stated
- AA-Omniscience · Component of the Artificial Analysis Intelligence Index; version unspecified
- ActivityNet-QA · Test subset
- AgentArena · Net Improvement;Qwen4-Max官方主图/正文;未注明的评测细节保持未知
- Agents’ Last Exam · 1.1 · 题集/指标未注明;Qwen4-Max官方主图/正文;未注明的评测细节保持未知
- Agents’ Last Exam · 55 professional fields; chart-reported accuracy; exact task version unspecified
- Agents’ Last Exam · 55 professional fields; exact task version unspecified
- Agents' Last Exam · GLM-5.3-Flash模型卡;unknown
- Agents' Last Exam · Metric and task revision unspecified in V4 cards
- Agents' Last Exam · Metric and task revision unspecified in V4.1-Flash-1017 card
- Agents' Last Exam · Task revision unspecified; Pass@1 as printed in V4.1-Flash card
- Agents’ Last Exam · V1; 55 subfields; exact task snapshot unspecified
- Agents' Last Exam · v1.1, 122 tasks released 2026-10-17 (before 2026-10-26 update)
- Agents' Last Exam · v1.1; current chart task snapshot and scoring details unspecified; not assumed to be the 2026-10-17 snapshot
- Agents' Last Exam · Version not stated in selected source
- Agents' Last Exam · 版本/单位未注明 (qg-completion-qwen38-27b-agents-last-exam)
- Agents' Last Exam · 版本/单位未注明 (qg-completion-qwen38-flash-next-agents-last-exam)
- Agents' Last Exam · 版本/单位未注明 (qg-completion-qwen38-max-api-agents-last-exam)
- Agents' Last Exam (ALE-CLI) · GLM-5.3模型卡;ALE-CLI
- AGIEval · DeepSeek-V3 Technical Report Table 3; AGIEval (EM); unit is not printed
- AI2D · Version not stated
- AI2D_TEST · Qwen3-VL-235B-A22B-Thinking model-card image table; Text Recognition and Chart/Document Understanding; label as printed; unit and metric definition not stated
- AI2D_TEST · 版本/单位未注明 (qg-completion-qwen35-ai2d-test)
- AI2D_TEST · 版本/单位未注明 (qg-completion-qwen36-ai2d-test)
- AI2D_with_Masks · Qwen3-VL技术报告v2;Table 2; AI2D multiple-choice prompt in Appendix B.4; sampling parameters not stated
- Aider Polyglot · Diff or diff-fenced edit format
- Aider-Edit · DeepSeek-V3 Technical Report Table 6; Aider-Edit (Acc.); unit not printed
- Aider-Polyglot · DeepSeek-R1 paper Table 3; Aider-Polyglot (Acc.); unit not printed
- Aider-Polyglot · Version unspecified; Accuracy
- Aider-Polyglot · 版本/单位未注明
- Aider-Polyglot Score (%) · Coder-Next技术报告;Polyglot · 版本未注明
- AIME · 2024 edition
- AIME · 2024 edition; Cons@64; table does not print a percent sign
- AIME · 2024 edition; Pass@1
- AIME · 2024 edition; Pass@1; table does not print a percent sign
- AIME · 2024; metric not printed in Qwen3-8B comparison; table does not print a percent sign
- AIME · 2025 edition
- AIME · 2025 edition; Pass@1
- AIME · 2025; metric not printed in Qwen3-8B comparison; table does not print a percent sign
- AIME · 2025; Pass@1; DeepSeek-R1 paper Table 13 and accompanying prose
- AIME · DeepSeek-R1 paper Table 3; 2024 (Pass@1); unit not printed
- AIME · DeepSeek-V3.2 report Table 2; 2025 (Pass@1); unit not printed
- AIME · DeepSeek-V3.2 report Table 3; Speciale explicitly named; 2025 (Pass@1); unit not printed
- AIME · GLM-5固定v2报告/模型卡;2026 I
- AIME · Year/metric not stated in V3-0324 overview
- AIME 2025 · GLM-4.7模型卡;2025; full I/II subset not stated
- AIME 2026 · GLM-5.1模型卡;2026; full I/II subset not stated
- AIME 2026 · GLM-5.2模型卡;2026; full I/II subset not stated
- AIME 25 · GLM-4.6模型卡;2025; full I/II subset not stated
- AIME 25 · GLM-4.6模型卡;2025; full I/II subset not stated; w/ Tools
- AIME24 · Coder-Next技术报告;版本/单位未注明
- AIME25 · 2025 · unit unspecified
- AIME25 · Qwen3-VL-235B-A22B-Thinking model-card image table; Reasoning; label as printed; unit and metric definition not stated
- AIME25 · 版本/单位未注明
- AIME26 · 2026 (I & II per footnote) · unit unspecified (qg-completion-qwen35-aime26)
- AIME26 · 2026 (I & II per footnote) · unit unspecified (qg-completion-qwen36-aime26)
- All-Angles Bench · Version not stated
- AlpacaEval 2.0 · DeepSeek-R1 paper Table 3; AlpacaEval 2.0 (LC-winrate); unit not printed
- AlpacaEval 2.0 · DeepSeek-V3 Technical Report Table 7; AlpacaEval 2.0 (length-controlled win rate); unit not printed
- AMC 12 · 2024; DeepSeek-R1 paper Table 13
- AndroidBench · 95-task public subset; avg@3; score scale not stated
- AndroidWorld · Qwen3-VL技术报告v2;Table 2; online environment agent benchmark; benchmark-specific prompt is not separately shown in Appendix B.11
- AndroidWorld · Version/protocol not stated
- AndroidWorld · 版本/单位未注明 (qg-completion-qwen35-androidworld)
- AndroidWorld · 版本/单位未注明 (qg-completion-qwen36-androidworld)
- AndroidWorld · 版本/单位未注明 (qg-completion-qwen38-27b-androidworld)
- AndroidWorld · 版本/单位未注明 (qg-completion-qwen38-flash-next-androidworld)
- Apex · DeepSeek-V4 report Table 7; MathArena Apex (Pass@1); unit not printed
- Apex Bench · v1.1; Pass@1 as printed in V4.1-Flash-1017 card
- Apex Shortlist · DeepSeek-V4 report Table 7; Apex Shortlist (Pass@1); unit not printed
- APEX-Agents · 1.1 · 指标未注明;Qwen4-Max官方主图/正文;未注明的评测细节保持未知
- APEX-Agents · Long-horizon professional tasks
- ARC-AGI-1 · ARC Prize Foundation private dataset
- ARC-AGI-1 · ARC Prize Foundation semi-private 2026 evaluation set
- ARC-AGI-1 · Verified set; release article does not specify task count
- ARC-AGI-1 · Verified; exact subset not specified
- ARC-AGI-1 · Version/subset unspecified in source
- ARC-AGI-2 · ARC Prize Foundation private dataset
- ARC-AGI-2 · ARC Prize Foundation semi-private 2026 evaluation set
- ARC-AGI-2 · ARC Prize Verified · semi-private set
- ARC-AGI-2 · Verified;题集组成未在发布表中说明
- ARC-AGI-2 · Version/subset unspecified in source
- ARC-AGI-3 · ARC Prize Foundation semi-private 2026 evaluation set
- ARC-AGI-3 · Responses API harness with two settings changed to better match real-world performance
- ARC-AGI-3 · Version/subset unspecified in source
- ARC-AGI-3 · 版本/子集未注明
- ARC-Challenge · DeepSeek-V3 Technical Report Table 3; ARC-Challenge (EM); unit is not printed
- ARC-Easy · DeepSeek-V3 Technical Report Table 3; ARC-Easy (EM); unit is not printed
- Arena-Hard · DeepSeek-R1 paper Table 3; Arena-Hard (GPT-4-1106); unit not printed
- Arena-Hard · DeepSeek-V3 Technical Report Table 7; Arena-Hard; unit not printed
- Arena-Hard v2 · 2 · GPT-4.1裁判 · 单位未注明
- Arena-Hard V2 · Qwen3-VL技术报告v2;Table 6; page 20 footnote says win rates evaluated by GPT-4.1; Thinking MoE sampling settings
- Arena-Hard v2 · 版本/单位未注明
- ARKitScenes · Qwen3-VL-235B-A22B-Thinking model-card image table; 2D/3D Grounding; label as printed; unit and metric definition not stated
- ARKitScenes · Qwen3-VL技术报告v2;Table 2; page 19 states Omni3D test, IoU threshold 0.15, confidence 1.0; 3D bbox JSON prompt in Appendix B.5
- Artificial Analysis Coding Agent Index · v1.1
- Artificial Analysis Coding Agent Index · v1.4
- Artificial Analysis Coding Index · Exact edition unspecified in article
- Artificial Analysis Intelligence Index · Exact edition unspecified in chart
- Artificial Analysis Intelligence Index · v4.1
- Artificial Analysis Intelligence Index · v4.1.1; source table does not state unit
- Artificial Analysis Intelligence Index · v5.0 with tools
- Artificial Analysis Intelligence Index · v5.0.1; nine evaluations
- Artificial Analysis Intelligence Index · Version not stated in selected table
- ArXivMath · April-May 2026 releases (81 problems)
- ArXivMath · August 2026 release (57 problems)
- ArXivMath · June 2026 release (49 problems)
- ArXivMath · March-April 2026 releases; 71 problems
- Automation-Bench · 600-task public subset; Pass@1; task release and unit not stated
- AutomationBench · 1.0.6; 47 tools; chart-reported score, scoring rule unspecified
- AutomationBench · GLM-5.3-Flash模型卡;v1.0.6
- AutomationBench · Public set; metric/version details not stated in V4-0813/0731 cards
- AutomationBench · Release 1.0.6, Zapier private evaluation set
- AutomationBench · V1.0.6 label; metric not stated in V4.1-Flash-1017 card
- AutomationBench · v1.0.6; Pass@1 as printed in V4.1-Flash card
- AutomationBench · Version unspecified; chart-reported accuracy
- AutomationBench · Zapier private set; release/version not stated
- AutomationBench · Zapier 私有 early-access set;版本未注明
- AutomationBench · 版本、任务集未在 Fable 发布表注明
- AutomationBench (v1.0.6) · GLM-5.3模型卡;v1.0.6
- AutomationBench-AA · AA · 版本/指标未注明;Qwen4-Max官方主图/正文;未注明的评测细节保持未知
B
- BabyVision · With tools; metric/task version unspecified in 1017 card
- BabyVision · With tools; task revision unspecified
- BabyVision · 版本/单位未注明 (qg-completion-qwen35-babyvision)
- BabyVision · 版本/单位未注明 (qg-completion-qwen38-27b-babyvision)
- BBH · DeepSeek-V3 Technical Report Table 3; BBH (EM); unit is not printed
- BenchCAD · 2.1 core scores
- BenchCAD · 2.1 Core; tools and geometric metric not specified in this chart
- BenchCAD · Task version and scoring details unspecified
- BenchCAD · Task version unspecified; Python tool protocol
- BenchCAD · Version unspecified; tool-assisted 3D reconstruction; geometric-overlap score
- BenchCAD Vision2Code · 1,000-file random subset of 17,900 files
- Berkeley Function Calling Leaderboard · v3
- BetterChartQA · 9 capability buckets
- BFCL · Qwen3-VL-235B-A22B-Thinking model-card image table; Agent; label as printed; unit and metric definition not stated
- BFCL-v3 · Qwen3-VL技术报告v2;Table 6; Thinking MoE sampling settings from pages 20-21
- BFCL-v3 · 版本/单位未注明 (qg-completion-qwen3-235b-2507-bfcl-v3)
- BFCL-v3 · 版本/单位未注明 (qs-completion-bfcl-v3)
- BFCL-V4 · 版本/单位未注明
- Big Finance Bench · Version and task protocol unspecified
- Biomedical Image Analysis · System-card evaluation across public biomedical-image competition tasks; included competition set unspecified
- BioMysteryBench · 官方发布表列 Hard 与 human-solved 结果;细分版本未注明
- BioMysteryBench · 版本/题集/指标未注明;Qwen4-Max官方主图/正文;未注明的评测细节保持未知
- BIRD-SQL · Coder-Next技术报告;版本/单位未注明
- BixBench · Version/subset unspecified
- BLINK · Qwen3-VL-235B-A22B-Thinking model-card image table; Multi-Image; label as printed; unit and metric definition not stated
- BLINK · Validation
- Blueprint-Bench · 2
- BrowseComp · 1,266 题;单代理与多代理按协议区分
- BrowseComp · DeepSeek-V3.2 report Table 2; BrowseComp (Pass@1); context-management result marked *; unit not printed
- BrowseComp · DeepSeek-V3.2 report Table 2; BrowseComp (Pass@1); without context management; unit not printed
- BrowseComp · GLM-4.5 report; protocol/version unspecified
- BrowseComp · GLM-4.7模型卡;unknown
- BrowseComp · GLM-5.1模型卡;unknown
- BrowseComp · GLM-5固定v2报告/模型卡;with context management
- BrowseComp · GLM-5固定v2报告/模型卡;without context management
- BrowseComp · Protocol/version unspecified in GLM source
- BrowseComp · unit unspecified
- BrowseComp · Version unspecified; Accuracy
- BrowseComp · Version unspecified; Pass@1
- BrowseComp (w/ Context Manage) · GLM-4.7模型卡;w/ Context Manage
- BrowseComp (w/ Context Manage) · GLM-5.1模型卡;w/ Context Manage
- BrowseComp Long Context · 128k context
- BrowseComp Long Context · 256k context
- BrowseComp-Zh · GLM-4.7模型卡;Zh
C
- CC_OCR · Qwen3-VL-235B-A22B-Thinking model-card image table; Text Recognition and Chart/Document Understanding; label as printed; unit and metric definition not stated
- CC-OCR · 版本/单位未注明 (qg-completion-qwen35-cc-ocr)
- CC-OCR · 版本/单位未注明 (qg-completion-qwen36-cc-ocr)
- CFEval · Qwen3-VL技术报告v2;Table 6; Thinking MoE sampling settings from pages 20-21
- CFEval · 版本/单位未注明 (qg-completion-qwen3-235b-2507-cfeval)
- CFEval · 版本/单位未注明 (qs-completion-cfeval)
- Charades-STA_mIoU · Qwen3-VL技术报告v2;Table 2; Charades-STA timestamp prompt in Appendix B.8; 4 fps; max 2,048 frames and 224K visual tokens
- CharadesSTA · Qwen3-VL-235B-A22B-Thinking model-card image table; Video; label as printed; unit and metric definition not stated
- ChartMimic · Qwen3-VL技术报告v2;Table 2; multimodal coding benchmark; specific prompt is not separately shown in Appendix B.10
- ChartMuseum · 1,162 test-split questions
- ChartMuseum · Set/version not stated
- Chartography · 100 tasks
- Chartography · no tools;task/version 未注明
- Chartography · Version/task count not stated in selected source
- Chartography · With tools; metric/task version unspecified in 1017 card
- Chartography · With tools; task revision unspecified
- ChartoGraphy · 题集/工具/单位未注明;Qwen4-Max官方主图/正文;未注明的评测细节保持未知
- ChartQA_test · Qwen3-VL技术报告v2;Table 2; ChartQA test single-word-or-phrase prompt in Appendix B.4; sampling parameters not stated
- ChartQAPro · ChartQAPro; set size and release version not stated in the System Card
- CharXiv (RQ) · 版本/单位未注明 (qg-completion-qwen38-27b-charxiv-rq)
- CharXiv (RQ) · 版本/单位未注明 (qg-completion-qwen38-flash-next-charxiv-rq)
- CharXiv Reasoning · Validation split
- CharXiv Reasoning · Version/split not stated in selected source
- CharXiv RQ · 版本/单位未注明
- CharXiv_DQ · Qwen3-VL技术报告v2;Table 2; CharXiv prompt in Appendix B.4; DQ description subset; sampling parameters not stated
- CharXiv(RQ) · Qwen3-VL-235B-A22B-Thinking model-card image table; Text Recognition and Chart/Document Understanding; label as printed; unit and metric definition not stated
- CharXiv(RQ) · 版本/单位未注明
- Claw-Eval Avg · 版本/单位未注明
- Claw-Eval Pass^3 · 版本/单位未注明
- ClawEval-MM · 版本/单位未注明 (qg-completion-qwen38-27b-claweval-mm)
- ClawEval-MM · 版本/单位未注明 (qg-completion-qwen38-flash-next-claweval-mm)
- CMO · 2025; overall points and gold medal; DeepSeek-V3.2 report Table 4
- CNMO · 2024 edition; Pass@1
- CNMO · DeepSeek-R1 paper Table 3; CNMO 2024 (Pass@1); unit not printed
- CNMO · DeepSeek-V3 Technical Report Table 6; CNMO 2024 (Pass@1); unit not printed
- CodeArena · 版本未注明;Qwen4-Max官方主图/正文;未注明的评测细节保持未知
- Codeforces · Codeforces;报告未说明指标定义、单位或任务版本。
- Codeforces · DeepSeek-R1 paper Table 3; Codeforces (Percentile); unit not printed
- Codeforces · DeepSeek-R1 paper Table 3; Codeforces (Rating); unit printed as rating
- Codeforces · DeepSeek-V3.2 report Table 2; Codeforces (Rating); unit printed as rating
- Codeforces · DeepSeek-V3.2 report Table 3; Speciale explicitly named; Codeforces (Rating); unit printed as rating
- Codeforces · Percentile; contest snapshot unspecified
- Codeforces · Rating; contest snapshot unspecified
- Codeforces · Rating; contest snapshot unspecified; table does not print a percent sign
- Codeforces-Div1 · Div1; contest snapshot unspecified
- Coder Aider-Polyglot · 题集/单位未注明
- Coder BFCL-v3 · 题集/单位未注明
- Coder Mind2Web · 题集/单位未注明
- Coder Multi-SWE-bench flash · 题集/单位未注明
- Coder Multi-SWE-bench mini · 题集/单位未注明
- Coder Spider2 · 题集/单位未注明
- Coder SWE-bench Live · 题集/单位未注明
- Coder SWE-bench Multilingual · 题集/单位未注明
- Coder TAU-Bench Airline · 题集/单位未注明
- Coder TAU-Bench Retail · 题集/单位未注明
- Coder Terminal-Bench · 版本/单位未注明
- Coder WebArena · 题集/单位未注明
- COLLIE · Version unspecified in GPT-5 developer release
- CorpusQA 1M · DeepSeek-V4 report Table 7; CorpusQA 1M (ACC); unit not printed
- CountBench · Qwen3-VL-235B-A22B-Thinking model-card image table; 2D/3D Grounding; label as printed; unit and metric definition not stated
- CountBench · 版本/单位未注明 (qg-completion-qwen35-countbench)
- CountBench · 版本/单位未注明 (qg-completion-qwen36-countbench)
- CoVoST2 · 21 languages to English
- Creative Writing v3 · Qwen3-VL-235B-A22B-Thinking model-card image table; Subjective Evaluation; label as printed; unit and metric definition not stated
- Creative Writing v3 · 版本/单位未注明 (qg-completion-qwen3-235b-2507-creative-writing-v3)
- Creative Writing v3 · 版本/单位未注明 (qs-completion-creative-writing-v3)
- CritPt · Component of the Artificial Analysis Intelligence Index; version unspecified
- CritPt · GLM-5.2模型卡;unknown
- CritPT-Corrected · 71 theoretical-physics problems; 31 statements corrected for underspecification, ambiguity, or typos; Pass@1 averaged over 16 attempts per problem
- CRUXEval · Coder-Next技术报告;版本/单位未注明
- CRUXEval-I · DeepSeek-V3 Technical Report Table 3; CRUXEval-I (EM); unit is not printed
- CRUXEval-O · DeepSeek-V3 Technical Report Table 3; CRUXEval-O (EM); unit is not printed
- CursorBench · 3.2.0
- CursorBench · 4.0
- CursorBench · Not stated in selected source
- CV-Bench · Version not stated
D
- De novo protein binder design · System Card evaluation; 15 targets; 30 miniprotein binders per target; Cas9 excluded
- DeepPlanning · 版本/单位未注明 (qg-completion-qwen35-deepplanning)
- DeepPlanning · 版本/单位未注明 (qg-completion-qwen36-deepplanning)
- DeepSearchQA · 900 prompts;单代理与多代理按协议区分
- DeepSWE · 1.1 · 指标未注明;Qwen4-Max官方主图/正文;未注明的评测细节保持未知
- DeepSWE · GLM-5.2模型卡;unknown
- DeepSWE · GLM-5.3-Flash模型卡;v1.1
- DeepSWE · v1.1
- DeepSWE · v1.1; metric not printed in V4.1-Flash-1017 card
- DeepSWE · v1.1; Resolved metric as printed in V4.1-Flash card
- DeepSWE · Version/metric unspecified in V4-0813/0731 cards
- DeepSWE (v1.1) · GLM-5.3模型卡;v1.1
- DeepSWE 1.1 · 版本/单位未注明 (qg-completion-qwen38-27b-deepswe-1-1)
- DeepSWE 1.1 · 版本/单位未注明 (qg-completion-qwen38-flash-next-deepswe-1-1)
- DeepSWE 1.1 · 版本/单位未注明 (qg-completion-qwen38-max-api-deepswe-1-1)
- Design2Code · Qwen3-VL-235B-A22B-Thinking model-card image table; Coding; label as printed; unit and metric definition not stated
- Design2Code · Version not stated
- DocVQA_TEST · Qwen3-VL-235B-A22B-Thinking model-card image table; Text Recognition and Chart/Document Understanding; label as printed; unit and metric definition not stated
- DRACO · 100-task Perplexity deep-research benchmark
- DROP · DeepSeek-R1 paper Table 3; DROP (3-shot F1); unit not printed
- DROP · DeepSeek-V3 Technical Report Table 3; DROP (F1); unit is not printed
- DROP · DeepSeek-V3 Technical Report Table 6; DROP (3-shot F1); unit not printed
- DROP · Version unspecified; 3-shot F1
- DynaMath · Qwen3-VL技术报告v2;Table 2; DynaMath JSON-answer prompt in Appendix B.1; sampling parameters not stated
- DynaMath · Version not stated
- DynaMath · 版本/单位未注明 (qg-completion-qwen35-dynamath)
- DynaMath · 版本/单位未注明 (qg-completion-qwen36-dynamath)
E
- ECLeKTic · Closed-book QA
- EgoTempo · Test subset
- EmbSpatialBench · Qwen3-VL-235B-A22B-Thinking model-card image table; Embedded and Spatial Understanding; label as printed; unit and metric definition not stated
- EmbSpatialBench · 版本/单位未注明 (qg-completion-qwen35-embspatialbench)
- EmbSpatialBench · 版本/单位未注明 (qg-completion-qwen36-embspatialbench)
- Epoch Capabilities Index · Pre-release ECI quoted in the 2026-11-17 GPT-6.2 system card; exact index snapshot/version unspecified; continually updated
- ERQA · Qwen3-VL-235B-A22B-Thinking model-card image table; Embedded and Spatial Understanding; label as printed; unit and metric definition not stated
- ERQA · Version not stated
- ERQA · Version unspecified in GPT-5 developer release
- ERQA · 版本/单位未注明 (qg-completion-qwen35-erqa)
- ERQA · 版本/单位未注明 (qg-completion-qwen36-erqa)
- ERQA · 版本/单位未注明 (qg-completion-qwen38-27b-erqa)
- ERQA · 版本/单位未注明 (qg-completion-qwen38-flash-next-erqa)
- EvalPlus · Coder-Next技术报告;版本/单位未注明
F
- FACTS Benchmark Suite · Held-out internal grounding, parametric, multimodal and search-retrieval benchmarks
- FACTS Grounding · 2024 release
- FActScore · GPT-5 developer release; no tools; hallucination-rate metric
- Finance Agent · Opus 4.6 来源未注明版本;发布于 Vals AI v2 (2026-05) 之前
- Finance Agent · v2
- Finance Agent · Vals AI SEC-filing research benchmark
- FinanceAgent · v1.1
- Flame-React-Eval · Version not stated
- FLEURS · 53-language subset
- FRAMES · DeepSeek-R1 paper Table 3; FRAMES (Acc.); unit not printed
- FRAMES · DeepSeek-V3 Technical Report Table 6; FRAMES (Acc.); unit not printed
- FRAMES · Version unspecified; Accuracy
- Frontier-Bench · v0.1
- FrontierCode · 1.1 Extended; chart-reported accuracy
- FrontierCode · 2.0
- FrontierCode · v1
- FrontierCode · v1.1 Extended set
- FrontierCode · v1.1 Main set
- FrontierMath · Set/version unspecified in GPT-5 and GPT-5.1 release tables; Python tool enabled.
- FrontierMath · Tier 1–3 v2
- FrontierMath · Tier 1–3; version not stated in 2025–2026 release article
- FrontierMath · Tier 1–3; version unspecified
- FrontierMath · Tier 4 v2
- FrontierMath · Tier 4 v2; chart-reported accuracy
- FrontierMath · Tier 4; version not stated in 2026-03 release article
- FrontierMath · Tier 4; version unspecified
- FrontierMath Erdős · 68 problems posed or studied by Erdős, open as of August 2026; exact evaluation revision unspecified
- FrontierScience · Research;版本与题集未注明
- FrontierSWE · MEAN@5 leaderboard score as of 2026-08-03; task release/version not stated
- FrontierSWE · Release/version unspecified in GLM-5.3 card
- FrontierSWE · v2
- FrontierSWE (Dominance) · GLM-5.2模型卡;Dominance
G
- GDM-MRCR · v2, 8-needle, 128k average context
- GDM-MRCR · v2, 8-needle, 1M pointwise context
- GDP.pdf · 100 professional PDFs and prompts
- GDP.pdf · OpenAI 2026-09 chart; 10 professional domains; chart score, scoring rule and exact set version unspecified
- GDP.PDF · Version not stated in selected source
- GDPval · 44 occupations; exact task set/version unspecified in release articles
- GDPval · 44 occupations; exact task/version unspecified
- GDPval · v2 internal; scoring rubric and exact task snapshot unspecified
- GDPval-AA · Artificial Analysis leaderboard
- GDPval-AA · DeepSeek-V4 report Table 7; GDPval-AA (Elo); unit printed as Elo
- GDPval-AA · GLM-5固定v2报告/模型卡;snapshot recorded 2026-02-15
- GDPval-AA · v2
- GDPval-AA · v2.1
- GDPval-AA v2 · GLM-5.3-Flash模型卡;v2
- GDPval-AA v2 · GLM-5.3模型卡;v2
- GeneBench · 2026-04-23 report; 103 evaluations across 10 domains
- GeneBench Pro · v13; chart-reported accuracy
- GeneBench Pro · v13; see official footnote in main article
- GeneBench Pro · Version unspecified in source
- GeoBench · Version not stated
- Global MMLU · 42 languages including English · overall average
- Global MMLU Lite · Lite; 15 human-translated languages
- Global PIQA · Commonsense reasoning across 100 languages and cultures
- Global PIQA · 版本/单位未注明
- GPQA · GLM-5固定v2报告/模型卡;Diamond
- GPQA · Official GLM-4.5 report; subset/version not stated
- GPQA · Qwen3-VL技术报告v2;Table 6; Thinking MoE sampling settings from pages 20-21
- GPQA · Subset/version unspecified in GLM source
- GPQA · V3-0324 overview does not state variant or metric
- GPQA · 版本/单位未注明 (qg-completion-qwen3-235b-2507-gpqa)
- GPQA · 版本/单位未注明 (qg-completion-qwen35-gpqa)
- GPQA · 版本/单位未注明 (qg-completion-qwen36-gpqa)
- GPQA · 版本/单位未注明 (qs-completion-gpqa)
- GPQA Diamond · 198-question Diamond subset
- GPQA Diamond · DeepSeek-R1 paper Table 3; GPQA Diamond (Pass@1); unit not printed
- GPQA Diamond · DeepSeek-V3.2 report Table 2; GPQA Diamond (Pass@1); unit not printed
- GPQA Diamond · DeepSeek-V3.2 report Table 3; Speciale explicitly named; GPQA Diamond (Pass@1); unit not printed
- GPQA Diamond · Diamond subset; task snapshot unspecified
- GPQA Diamond · Diamond subset; task snapshot unspecified; table does not print a percent sign
- GPQA Diamond · Variant/metric not printed in Qwen3-8B comparison; table does not print a percent sign
- GPQA Diamond · 版本/单位未注明 (qg-completion-qwen36-gpqa-diamond)
- GPQA Diamond · 版本/单位未注明 (qg-completion-qwen38-27b-gpqa-diamond)
- GPQA Diamond · 版本/单位未注明 (qg-completion-qwen38-flash-next-gpqa-diamond)
- GPQA Diamond · 版本/单位未注明 (qg-completion-qwen38-max-api-gpqa-diamond)
- GPQA-Diamond · GLM-4.7模型卡;Diamond
- GPQA-Diamond · GLM-5.1模型卡;Diamond
- GPQA-Diamond · GLM-5.2模型卡;Diamond
- GraphWalks · 200 problems, 256k to 1M context, BFS F1
- GraphWalks · 256K-context subset of the 256K-1M evaluation; item count unspecified; BFS F1
- GraphWalks · 256K-context subset of the 256K-1M evaluation; item count unspecified; parents
- GraphWalks · 650 items, context up to 128k, BFS F1
- GraphWalks · BFS;0–128K 上下文;题数未注明
- GraphWalks · BFS;256K–1M 上下文;题数未注明
- GraphWalks · parents;0–128K 上下文;题数未注明
- GraphWalks · parents;256K–1M 上下文;题数未注明
- GSM8K · DeepSeek-V3 Technical Report Table 3; GSM8K (EM); unit is not printed
H
- HallusionBench · Qwen3-VL-235B-A22B-Thinking model-card image table; Subjective Experience and Instruction Following; label as printed; unit and metric definition not stated
- HallusionBench · Version not stated
- HallusionBench · 版本/单位未注明 (qg-completion-qwen35-hallusionbench)
- HallusionBench · 版本/单位未注明 (qg-completion-qwen36-hallusionbench)
- Harvey LAB-AA · Artificial Analysis v1.1; 120-task held-out set; hallucination-gated and ungated all-pass rates
- Harvey LAB-AA · Version not stated in selected source
- HealthBench · 48,000+ rubric items across 5,000 multi-turn conversations
- HealthBench · Task version, score variant, and metric scale not stated in the comparison table
- HealthBench Consensus · System-card evaluation; exact dataset version unspecified
- HealthBench Hard · System-card evaluation; exact dataset version unspecified
- HealthBench Professional · 525 physician-authored conversations
- HealthBench Professional · 校正口径/单位未注明;Qwen4-Max官方主图/正文;未注明的评测细节保持未知
- HellaSwag · DeepSeek-V3 Technical Report Table 3; HellaSwag (EM); unit is not printed
- HiddenMath-Hard · Hard held-out set
- HLAB-Vals · 1.1 · 指标未注明;Qwen4-Max官方主图/正文;未注明的评测细节保持未知
- HLE · DeepSeek-V3.2 report Table 2; HLE (Pass@1); reasoning/no-tools table row; unit not printed
- HLE · DeepSeek-V3.2 report Table 2; HLE (Pass@1); Search Agent table section; unit not printed
- HLE · DeepSeek-V3.2 report Table 3; Speciale explicitly named; HLE (Pass@1); unit not printed
- HLE · GLM-4.7模型卡;unknown; full/text-only subset not stated
- HLE · GLM-5.1模型卡;unknown; full/text-only subset not stated
- HLE · GLM-5.2模型卡;text-only; full set not indicated
- HLE · HLE label only; question-set scope, metric and unit not stated in the cited table
- HLE · HLE with tools; question-set scope, metric and unit not stated in the cited table
- HLE · Official HLE template; V3.2 thinking; unit not printed
- HLE · 版本/单位未注明 (qg-completion-qwen3-235b-2507-hle)
- HLE · 版本/单位未注明 (qg-completion-qwen35-hle)
- HLE · 版本/单位未注明 (qg-completion-qwen36-hle)
- HLE · 版本/单位未注明 (qg-completion-qwen38-27b-hle)
- HLE · 版本/单位未注明 (qg-completion-qwen38-flash-next-hle)
- HLE (w/ Tools) · GLM-4.7模型卡;w/ Tools; full/text-only subset not stated
- HLE (w/ Tools) · GLM-5.1模型卡;w/ Tools; full/text-only subset not stated
- HLE (w/ Tools) · GLM-5.2模型卡;text-only; w/ Tools; full set not indicated
- HLE w/ tool · 版本/单位未注明
- HLE w/ tools · DeepSeek-V4 report Table 7; HLE w/ tools (Pass@1); unit not printed
- HLE w/ Tools · GLM-5.3-Flash模型卡;full set; w/ Tools
- HLE w/ Tools · GLM-5.3模型卡;w/ Tools; full/text-only subset not stated
- HLE-Verified¹ · 版本/单位未注明
- HMMT · 2025 competition; Pass@1
- HMMT · 2025 edition; reported score, metric not stated in V3.2-Exp card
- HMMT · DeepSeek-V3.2 report Table 2; February 2025 (Pass@1); unit not printed
- HMMT · DeepSeek-V3.2 report Table 2; November 2025 (Pass@1); unit not printed
- HMMT · DeepSeek-V3.2 report Table 3; Speciale explicitly named; February 2025 (Pass@1); unit not printed
- HMMT · DeepSeek-V3.2 report Table 3; Speciale explicitly named; November 2025 (Pass@1); unit not printed
- HMMT · DeepSeek-V4 report Table 7; 2026 February (Pass@1); unit not printed
- HMMT · February 2025; metric not stated in Qwen3-8B table; table does not print a percent sign
- HMMT · GLM-5固定v2报告/模型卡;Feb. 2025
- HMMT · GLM-5固定v2报告/模型卡;Nov. 2025
- HMMT 2025 · 2025 edition; no-tools protocol from GPT-5 developer release
- HMMT Feb 25 · February 2025 · unit unspecified (qg-completion-qwen35-hmmt-feb-25)
- HMMT Feb 25 · February 2025 · unit unspecified (qg-completion-qwen36-hmmt-feb-25)
- HMMT Feb 26 · February 2026 · unit unspecified
- HMMT Feb. 2025 · GLM-4.7模型卡;Feb. 2025
- HMMT Feb. 2026 · GLM-5.1模型卡;Feb. 2026
- HMMT Feb. 2026 · GLM-5.2模型卡;Feb. 2026
- HMMT Nov 25 · November 2025 · unit unspecified (qg-completion-qwen35-hmmt-nov-25)
- HMMT Nov 25 · November 2025 · unit unspecified (qg-completion-qwen36-hmmt-nov-25)
- HMMT Nov. 2025 · GLM-4.7模型卡;Nov. 2025
- HMMT Nov. 2025 · GLM-5.1模型卡;Nov. 2025
- HMMT Nov. 2025 · GLM-5.2模型卡;Nov. 2025
- HMMT25 · Qwen3-VL-235B-A22B-Thinking model-card image table; Reasoning; label as printed; unit and metric definition not stated
- HMMT25 · 版本/单位未注明 (qg-completion-qwen3-235b-2507-hmmt25)
- HMMT25 · 版本/单位未注明 (qs-completion-hmmt25)
- HMMT25 Feb · Coder-Next技术报告;版本/单位未注明
- HMMT25 Nov · Coder-Next技术报告;版本/单位未注明
- HRBench4K · Qwen3-VL技术报告v2;Table 2; thinking target has no + marker, so no tool use is indicated; HRBench tool prompt documented in Appendix B.9
- HRBench8K · Qwen3-VL技术报告v2;Table 2; thinking target has no + marker, so no tool use is indicated; HRBench tool prompt documented in Appendix B.9
- HumanEval · DeepSeek-V3 Technical Report Table 3; HumanEval (Pass@1); unit is not printed
- HumanEval-Mul · DeepSeek-V3 Technical Report Table 6; HumanEval-Mul (Pass@1); unit not printed
- Humanity's Last Exam · 2,500-question benchmark
- Humanity's Last Exam · DeepSeek V4 preview table; subset/version unspecified
- Humanity's Last Exam · Full HLE set as distinguished from the card-marked text-only subset; task revision unspecified
- Humanity's Last Exam · GLM-5固定v2报告/模型卡;text-only · w/Tools
- Humanity's Last Exam · GLM-5固定v2报告/模型卡;text-only subset
- Humanity's Last Exam · HLE-Verified 1,811-item verified set
- Humanity's Last Exam · no tools;题集版本未注明
- Humanity's Last Exam · R1-0528 card: text-only prompts; task revision unspecified
- Humanity's Last Exam · Release/task set unspecified; percentage explicitly stated in GLM-4.7 prose
- Humanity's Last Exam · Text questions; exact release unspecified
- Humanity's Last Exam · Text-only subset as stated in V3.1 and later comparison cards; task revision unspecified
- Humanity's Last Exam · Text-only subset marked † in DeepSeek V4.1-Flash card; task revision unspecified
- Humanity’s Last Exam · Tool-enabled protocol; exact subset/version unspecified
- Humanity's Last Exam · V4-0813/0731 card does not specify metric or task subset
- Humanity's Last Exam · V4.1-Flash-1017 comparison table; non-text-only value, metric/set unspecified
- Humanity's Last Exam · V4.1-Flash-1017 comparison table; text-only † value, metric/set unspecified
- Humanity's Last Exam · Version, tools and unit unspecified in chart
- Humanity's Last Exam · with tools;题集版本未注明
- Humanity's Last Exam · With-tools evaluation; task subset unspecified
- Humanity’s Last Exam · 题集/工具/单位未注明;Qwen4-Max官方主图/正文;未注明的评测细节保持未知
- Hypersim · Qwen3-VL-235B-A22B-Thinking model-card image table; 2D/3D Grounding; label as printed; unit and metric definition not stated
- Hypersim · Qwen3-VL技术报告v2;Table 2; page 19 states Omni3D test, IoU threshold 0.15, confidence 1.0; 3D bbox JSON prompt in Appendix B.5
- Hypersim · 版本/单位未注明
I
- ICPC WF · 2025; overall submissions/solved count and gold medal; DeepSeek-V3.2 report Table 4
- IF-Eval · DeepSeek-R1 paper Table 3; IF-Eval (PromptStrict); unit not printed
- IF-Eval · DeepSeek-V3 Technical Report Table 6; IF-Eval (PromptStrict); unit not printed
- IFBench · Component of the Artificial Analysis Intelligence Index; version unspecified
- IFBench · GLM-5固定v2报告/模型卡;版本未注明
- IFBench · 版本/单位未注明 (qg-completion-qwen35-ifbench)
- IFBench · 版本/单位未注明 (qg-completion-qwen38-27b-ifbench)
- IFBench · 版本/单位未注明 (qg-completion-qwen38-flash-next-ifbench)
- IFBench · 版本/单位未注明 (qg-completion-qwen38-max-api-ifbench)
- IFEval · GLM-4.5 Technical Report Table 6;精确题集修订未注明,单位未注明;指标按§4.2.4原文。
- IFEval · Qwen3-VL-235B-A22B-Thinking model-card image table; Instruction Following; label as printed; unit and metric definition not stated
- IFEval · 版本/单位未注明 (qg-completion-qwen3-235b-2507-ifeval)
- IFEval · 版本/单位未注明 (qg-completion-qwen35-ifeval)
- IFEval · 版本/单位未注明 (qs-completion-ifeval)
- IMO · 2025 competition; categorical medal result
- IMO · 2025; overall points and gold medal; DeepSeek-V3.2 report Table 4
- IMO-AnswerBench · GLM-5固定v2报告/模型卡;版本未注明
- IMOAnswerBench · DeepSeek-V3.2 report Table 2; IMOAnswerBench (Pass@1); unit not printed
- IMOAnswerBench · DeepSeek-V3.2 report Table 3; Speciale explicitly named; IMOAnswerBench (Pass@1); unit not printed
- IMOAnswerBench · DeepSeek-V4 report Table 7; IMOAnswerBench (Pass@1); unit not printed
- IMOAnswerBench · GLM-4.7模型卡;unknown
- IMOAnswerBench · GLM-5.1模型卡;unknown
- IMOAnswerBench · GLM-5.2模型卡;unknown
- IMOAnswerBench · 版本/单位未注明 (qg-completion-qwen35-imoanswerbench)
- IMOAnswerBench · 版本/单位未注明 (qg-completion-qwen36-imoanswerbench)
- INCLUDE · 44 languages
- InfoVQA_TEST · Qwen3-VL-235B-A22B-Thinking model-card image table; Text Recognition and Chart/Document Understanding; label as printed; unit and metric definition not stated
- International Mathematical Olympiad · 2026 contest; six proof problems
- IOI · 2025 competition; categorical medal result
- IOI · 2025; overall points and gold medal; DeepSeek-V3.2 report Table 4
J
L
- LAB-Bench FigQA · FutureHouse FigQA; prompting and tools are recorded per result
- LABBench2 · Version not stated in selected source
- LCB v6 · Qwen3-VL-235B-A22B-Thinking model-card image table; Reasoning; label as printed; unit and metric definition not stated
- Legal Agent Benchmark · Opus 5 官方发布表所列 held-out set
- Legal Agent Benchmark · Public full set
- LifeSciBench · Gold v1
- LifeSciBench · Version unspecified in source
- LifeVal · Version and exact task set unspecified in release article
- LingoQA · 版本/单位未注明
- LiveBench 2024-11-25 · Qwen3-VL技术报告v2;Table 6; Thinking MoE sampling settings from pages 20-21
- LiveBench 20241125 · 2024-11-25 · unit unspecified
- LiveBench 20241125 · 2024.11.25 · 单位未注明
- LiveBench1125 · Qwen3-VL-235B-A22B-Thinking model-card image table; Reasoning; label as printed; unit and metric definition not stated
- LiveCodeBench · 2408-2505 label; metric not stated in Qwen3-8B table; table does not print a percent sign
- LiveCodeBench · 2408-2505; Pass@1
- LiveCodeBench · DeepSeek-R1 paper Table 3; LiveCodeBench (Pass@1-COT); unit not printed
- LiveCodeBench · DeepSeek-V3 Technical Report Table 6; LiveCodeBench (Pass@1); unit not printed
- LiveCodeBench · DeepSeek-V3.2 report Table 2; LiveCodeBench (Pass@1-COT); unit not printed
- LiveCodeBench · DeepSeek-V3.2 report Table 3; Speciale explicitly named; LiveCodeBench (Pass@1-COT); unit not printed
- LiveCodeBench · Difficulty subset Easy; DeepSeek-R1 paper Table 14; unit not printed
- LiveCodeBench · Difficulty subset Hard; DeepSeek-R1 paper Table 14; unit not printed
- LiveCodeBench · Difficulty subset Medium; DeepSeek-R1 paper Table 14; unit not printed
- LiveCodeBench · Problems from 2024-07-01 through 2025-01-01
- LiveCodeBench · UI 2025-01-01 to 2025-05-01
- LiveCodeBench · v6; GLM source metric unit unspecified
- LiveCodeBench · Version unspecified in V4 preview table; Pass@1
- LiveCodeBench · Version unspecified; Pass@1
- LiveCodeBench · Version unspecified; Pass@1-CoT
- LiveCodeBench · Version unspecified; Pass@1; table does not print a percent sign
- LiveCodeBench · Version/metric not stated in V3-0324 overview
- LiveCodeBench Pro · Competitive coding problems from Codeforces, ICPC, and IOI
- LiveCodeBench v6 · Coder-Next技术报告;v6 · 时间题集/单位未注明
- LiveCodeBench v6 · Qwen3-VL技术报告v2;Table 6; Thinking MoE sampling settings; max output length 81,920 tokens
- LiveCodeBench v6 · v6 · benchmark release date/period not stated · unit unspecified (qg-completion-qwen35-livecodebench-v6)
- LiveCodeBench v6 · v6 · benchmark release date/period not stated · unit unspecified (qg-completion-qwen36-livecodebench-v6)
- LiveCodeBench v6 · v6 · benchmark release date/period not stated · unit unspecified (qg-completion-qwen38-27b-livecodebench-v6)
- LiveCodeBench v6 · v6 · benchmark release date/period not stated · unit unspecified (qg-completion-qwen38-flash-next-livecodebench-v6)
- LiveCodeBench v6 (25.02-25.05) · v6 · 2025.02–05 · 单位未注明
- LiveCodeBench v6 (25.02-25.05) · v6 · benchmark release date/period not stated · unit unspecified
- LiveCodeBench-Base · DeepSeek-V3 Technical Report Table 3; LiveCodeBench-Base 0801-1101 (Pass@1); unit is not printed
- LMArena · Gemini 3 release leaderboard score
- LMArena Code · GLM-5固定v2报告/模型卡;报告快照日期未注明
- LMArena Text · GLM-5固定v2报告/模型卡;报告快照日期未注明
- LOFT · Hard retrieval; 1M context, pointwise value
- LOFT · Hard retrieval; average context length up to 128K; 300 queries
- LogicVista · Qwen3-VL技术报告v2;Table 2; LogicVista prompt in Appendix B.1; sampling parameters not stated
- LogicVista · Version not stated
- LongBench · GLM-5固定v2报告/模型卡;v2
- LongBench v2 · DeepSeek-V3 Technical Report Table 6; LongBench v2 (Acc.); unit not printed
- LongBench v2 · Task version and metric scale not stated in the comparison table
- LongBench v2 · 版本/单位未注明
- LongFact-Concepts · GPT-5 developer release; no tools
- LongFact-Objects · GPT-5 developer release; no tools
- LVBench · Qwen3-VL-235B-A22B-Thinking model-card image table; Video; label as printed; unit and metric definition not stated
- LVBench · Test subset
- LVBench · Version not stated
- LVBench · 版本/单位未注明 (qg-completion-qwen35-lvbench)
- LVBench · 版本/单位未注明 (qg-completion-qwen36-lvbench)
- LVBench · 版本/单位未注明 (qg-completion-qwen38-flash-next-lvbench)
M
- MATH · DeepSeek-V3 Technical Report Table 3; MATH (EM); unit is not printed
- MATH-500 · 500-question set; Pass@1
- MATH-500 · 500-question set; Pass@1; table does not print a percent sign
- MATH-500 · DeepSeek-R1 paper Table 3; MATH-500 (Pass@1); unit not printed
- MATH-500 · DeepSeek-V3 Technical Report Table 6; MATH-500 (EM); unit not printed
- Math-VR · Qwen3-VL技术报告v2;Table 2; Math-VR result row; benchmark-specific prompt is not separately labeled in Appendix B
- MathArena Apex · MathArena leaderboard
- MathArena Apex · Version unspecified; Pass@1 as printed in V4.1-Flash card
- MathVerse · Version not stated
- MathVerse_mini · Qwen3-VL-235B-A22B-Thinking model-card image table; STEM&Puzzle; label as printed; unit and metric definition not stated
- MathVision · Qwen3-VL-235B-A22B-Thinking model-card image table; STEM&Puzzle; label as printed; unit and metric definition not stated
- MathVision · Version not stated
- MathVision · 版本/单位未注明 (qg-completion-qwen35-mathvision)
- MathVision · 版本/单位未注明 (qg-completion-qwen38-27b-mathvision)
- MathVision · 版本/单位未注明 (qg-completion-qwen38-flash-next-mathvision)
- MathVision-Wildphoto · Qwen3-VL技术报告v2;Table 2; MathVision-family prompt in Appendix B.1; sampling parameters not stated
- MathVista · Version not stated
- MathVista mini · 版本/单位未注明
- MathVista_mini · Qwen3-VL-235B-A22B-Thinking model-card image table; STEM&Puzzle; label as printed; unit and metric definition not stated
- Mathvista(mini) · 版本/单位未注明
- MAXIFE · 版本/单位未注明
- MBPP · DeepSeek-V3 Technical Report Table 3; MBPP (Pass@1); unit is not printed
- MCP Atlas · Public set
- MCP-Atlas · GLM-5固定v2报告/模型卡;500-task public set
- MCP-Atlas · Scale AI 公共排行榜;2026-04 刷新 harness
- MCP-Atlas · 所选材料未明确子集
- MCP-Atlas · 版本/单位未注明
- MCP-Atlas (Public Set) · GLM-5.1模型卡;Public Set
- MCP-Atlas (Public Set) · GLM-5.2模型卡;Public Set
- MCP-Mark · DeepSeek-V3.2 report Table 2; MCP-Mark (Pass@1); unit not printed
- MCP-Mark · 版本/单位未注明
- MCP-Universe · DeepSeek-V3.2 report Table 2; MCP-Universe (SuccessRate); unit not printed
- MCPAtlas Public · DeepSeek-V4 report Table 7; MCPAtlas Public (Pass@1); unit not printed
- MCPMark · 版本/单位未注明
- MedCalc-Bench · Verified dataset
- Medicinal Chemistry ADME · 504 drug-discovery questions about measured ADME properties
- MedXpertQA-MM · 版本/单位未注明
- MentalHealthBench · Overall; 1,215 synthetic conversations; four independent responses per task; clinician-authored rubrics
- MGSM · DeepSeek-V3 Technical Report Table 3; MGSM (EM); unit is not printed
- MIABench · Qwen3-VL-235B-A22B-Thinking model-card image table; Subjective Experience and Instruction Following; label as printed; unit and metric definition not stated
- Mind2Web · Benchmark version unspecified; updated Codex harness
- Minerva · Video; test subset
- MLE-Bench · 30 competition subset selected from tasks with <50GB data and <10h reference duration
- MLE-Bench · Canonical Partial 30 subset
- MLS-Bench-Lite · Task version and metric scale not stated in the comparison table
- MLVU · Qwen3-VL-235B-A22B-Thinking model-card image table; Video; label as printed; unit and metric definition not stated
- MLVU · 版本/单位未注明
- MLVU (M-Avg) · 版本/单位未注明
- MLVU_M-Avg · Qwen3-VL技术报告v2;Table 2; thinking-model video multiple-choice prompt in Appendix B.8; max 2,048 frames and 224K visual tokens; 2 fps
- MM_MT_Bench · Qwen3-VL-235B-A22B-Thinking model-card image table; Subjective Experience and Instruction Following; label as printed; unit and metric definition not stated
- MMBench · V1.1
- MMBrowseComp · Version not stated
- MMLongBench · 128K
- MMLongBench · Doc
- MMLongBench-Doc · Qwen3-VL-235B-A22B-Thinking model-card image table; Text Recognition and Chart/Document Understanding; label as printed; unit and metric definition not stated
- MMLongBench-Doc · Qwen3-VL技术报告v2;Table 2; page 17 text identifies the thinking value as 56.2%; multi-image prompt in Appendix B.4
- MMLongBench-Doc · 版本/单位未注明
- MMLU · Coder-Next技术报告;版本/单位未注明
- MMLU · DeepSeek-R1 paper Table 3; MMLU (EM); unit not printed
- MMLU · DeepSeek-V3 Technical Report Table 3; MMLU (EM); unit is not printed
- MMLU · DeepSeek-V3 Technical Report Table 6; MMLU (EM); unit not printed
- MMLU · GLM-4.5 Technical Report Table 6;精确题集修订未注明,单位未注明;指标按§4.2.4原文。
- MMLU · Qwen3-VL-235B-A22B-Thinking model-card image table; Knowledge; label as printed; unit and metric definition not stated
- MMLU · Version unspecified; Pass@1
- MMLU-Pro · DeepSeek card version unspecified; Exact Match
- MMLU-Pro · DeepSeek-R1 paper Table 3; MMLU-Pro (EM); unit not printed
- MMLU-Pro · DeepSeek-V3 Technical Report Table 3; MMLU-Pro (EM); unit is not printed
- MMLU-Pro · DeepSeek-V3.2 report Table 2; MMLU-Pro (EM); unit not printed
- MMLU-Pro · GLM-4.7模型卡;unknown
- MMLU-Pro · Qwen3-VL-235B-A22B-Thinking model-card image table; Knowledge; label as printed; unit and metric definition not stated
- MMLU-Pro · V3-0324 overview does not state metric or set version
- MMLU-Pro · 版本/单位未注明 (qg-completion-qwen3-235b-2507-mmlu-pro)
- MMLU-Pro · 版本/单位未注明 (qg-completion-qwen35-mmlu-pro)
- MMLU-Pro · 版本/单位未注明 (qg-completion-qwen36-mmlu-pro)
- MMLU-Pro · 版本/单位未注明 (qs-completion-mmlu-pro)
- MMLU-Redux · DeepSeek-R1 paper Table 3; MMLU-Redux (EM); unit not printed
- MMLU-Redux · DeepSeek-V3 Technical Report Table 3; MMLU-Redux (EM); unit is not printed
- MMLU-Redux · DeepSeek-V3 Technical Report Table 6; MMLU-Redux (EM); unit not printed
- MMLU-Redux · Qwen3-VL-235B-A22B-Thinking model-card image table; Knowledge; label as printed; unit and metric definition not stated
- MMLU-Redux · Version unspecified; Exact Match
- MMLU-Redux · 版本/单位未注明 (qg-completion-qwen3-235b-2507-mmlu-redux)
- MMLU-Redux · 版本/单位未注明 (qg-completion-qwen35-mmlu-redux)
- MMLU-Redux · 版本/单位未注明 (qg-completion-qwen36-mmlu-redux)
- MMLU-Redux · 版本/单位未注明 (qs-completion-mmlu-redux)
- MMMLU · 57 subjects and 14 non-English languages
- MMMLU · Multilingual Q&A
- MMMLU · 版本/单位未注明
- MMMU · Qwen3-VL技术报告v2;Table 2; MMMU prompt in Appendix B.1; sampling parameters not stated
- MMMU · Single-attempt
- MMMU · Validation split
- MMMU · 版本/单位未注明 (qg-completion-qwen35-mmmu)
- MMMU · 版本/单位未注明 (qg-completion-qwen36-mmmu)
- MMMU_Pro · Qwen3-VL-235B-A22B-Thinking model-card image table; STEM&Puzzle; label as printed; unit and metric definition not stated
- MMMU_VAL · Qwen3-VL-235B-A22B-Thinking model-card image table; STEM&Puzzle; label as printed; unit and metric definition not stated
- MMMU-Pro · Standard and Vision formats
- MMMU-Pro · Standard/Vision 等格式未在所选材料中注明
- MMMU-Pro · Tool-enabled protocol; exact version unspecified (gpt55-mmmu-pro-with-tools)
- MMMU-Pro · Tool-enabled protocol; exact version unspecified (gpt56-mmmu-pro-with-tools)
- MMMU-Pro · Vision format explicitly labeled by the official comparison chart; task release/version unspecified
- MMMU-Pro · 版本/单位未注明 (qg-completion-qwen35-mmmu-pro)
- MMMU-Pro · 版本/单位未注明 (qg-completion-qwen36-mmmu-pro)
- MMStar · Qwen3-VL-235B-A22B-Thinking model-card image table; General VQA; label as printed; unit and metric definition not stated
- MMStar · Version not stated
- MMStar · 版本/单位未注明 (qg-completion-qwen35-mmstar)
- MMStar · 版本/单位未注明 (qg-completion-qwen36-mmstar)
- MMVU · Qwen3-VL技术报告v2;Table 2; MMVU prompts in Appendix B.8; max 768 tokens/frame, 2,048 frames, 224K visual tokens; 2 fps
- MMVU · Version not stated
- MMVU · 版本/单位未注明
- Morphology-to-molecule matching · Four multi-channel human-liver in-vitro cell imaging sets matched to shuffled molecule-dose perturbations
- MotionBench · Version not stated
- MRCR 1M · DeepSeek-V4 report Table 7; MRCR 1M (MMR); unit not printed
- MRCR-V2 · 8-needle; pointwise at 1M context
- MRCR-V2 · 8-needle; up to 128K average context
- MUIRBENCH · Qwen3-VL-235B-A22B-Thinking model-card image table; Multi-Image; label as printed; unit and metric definition not stated
- MUIRBENCH · Version not stated
- Multi Challenge · GLM-5固定v2报告/模型卡;版本未注明
- Multi-task Indic Language Understanding Benchmark · 10 Indic languages plus English · overall average
- MultiChallenge · GLM-4.5 Technical Report Table 6;精确题集修订未注明,单位未注明;指标按§4.2.4原文。
- MultiChallenge · 版本/单位未注明
- MultiIF · 版本/单位未注明 (qg-completion-qwen3-235b-2507-multiif)
- MultiPL-E · 版本/单位未注明
- MVBench · Qwen3-VL技术报告v2;Table 2; thinking-model video multiple-choice prompt in Appendix B.8; max 2,048 frames and 224K visual tokens
- MVBench · Version not stated
- MVBench · 版本/单位未注明 (qg-completion-qwen35-mvbench)
- MVBench · 版本/单位未注明 (qg-completion-qwen36-mvbench)
N
- NaturalQuestions · DeepSeek-V3 Technical Report Table 3; NaturalQuestions (EM); unit is not printed
- Neptune · Video; test subset
- NL2Repo · GLM-5.1模型卡;unknown
- NL2Repo · GLM-5.2模型卡;unknown
- NL2Repo · GLM-5.3模型卡;unknown
- NL2Repo · Version/metric unspecified in V4-0813/0731 cards
- NL2Repo · 版本/单位未注明
- NL2Repo-Bench · Metric/task version unspecified in 1017 card
- NL2Repo-Bench · Task version and metric scale not stated in the comparison table
- NL2Repo-Bench · Version unspecified; Score metric as printed in V4.1-Flash card
- NL2Repo-Bench · 版本/单位未注明 (qg-completion-qwen38-27b-nl2repo-bench)
- NL2Repo-Bench · 版本/单位未注明 (qg-completion-qwen38-flash-next-nl2repo-bench)
- NOVA-63 · 版本/单位未注明
- Nuscene · 版本/单位未注明
O
- Objectron · Qwen3-VL-235B-A22B-Thinking model-card image table; 2D/3D Grounding; label as printed; unit and metric definition not stated
- OCR-Bench · v2 Chinese
- OCR-Bench · v2 English
- OCRBench · Qwen3-VL-235B-A22B-Thinking model-card image table; Text Recognition and Chart/Document Understanding; label as printed; unit and metric definition not stated
- OCRBench · Version not stated
- OCRBench · 版本/单位未注明 (qg-completion-qwen35-ocrbench)
- OCRBench · 版本/单位未注明 (qg-completion-qwen36-ocrbench)
- ODinW-13 · Qwen3-VL技术报告v2;Table 2; page 19 states mAP with confidence 1.0 and all dataset categories in the prompt; no %
- ODinW13 · Qwen3-VL-235B-A22B-Thinking model-card image table; 2D/3D Grounding; label as printed; unit and metric definition not stated
- ODInW13 · 版本/单位未注明 (qg-completion-qwen35-odinw13)
- ODInW13 · 版本/单位未注明 (qg-completion-qwen36-odinw13)
- OfficeQA · Full set
- OfficeQA · Pro 133-question subset
- OfficeQA · 题集与精确计分协议未注明
- OJBench · Qwen3-VL技术报告v2;Table 6; Thinking MoE sampling settings from pages 20-21
- OJBench · 版本/单位未注明 (qg-completion-qwen3-235b-2507-ojbench)
- OJBench · 版本/单位未注明 (qs-completion-ojbench)
- OmniDocBench · 版本未注明;归一化编辑距离
- OmniDocBench 1.5 · Version 1.5; average edit distance across text, formula, table and reading order
- OmniDocBench 1.5 · 版本/单位未注明
- OmniDocBench1.5 · 版本/单位未注明 (qg-completion-qwen35-omnidocbench1-5)
- OmniDocBench1.5 · 版本/单位未注明 (qg-completion-qwen36-omnidocbench1-5)
- OmniSpatial · Version not stated
- OpenAI MRCR · 2 needles; 128k context
- OpenAI MRCR · 2 needles; 256k context
- OpenAI MRCR · v2, 1M context, 8 needles
- OpenAI MRCR · v2, 256K context, 8 needles
- OpenAI MRCR v2 · v2 fix dated 2025-12-05; 8 needles; prompts in (128K, 256K] tokens; Mean Match Ratio
- OpenAI MRCR v2 · v2 fix dated 2025-12-05; 8 needles; prompts in (524K, 1024K] tokens; full variant may exceed the public API context limit
- OpenRCA · 335 enterprise software failure cases
- OpenScore String Quartets · OMR-NED; version unspecified
- Organic Chemistry · V2; system-card evaluation with expert-recommended harder problems
- OSWorld · 2.0 offline subset; partial score
- OSWorld · 2.0 v2026.08.08 offline set; partial reward score
- OSWorld · 2.0 v2026.08.08 offline subset; partial score
- OSWorld · 2.0,August 2026 task release(与此前 task files 不可直接比较)
- OSWorld · 2.0; partial score; 2026-08-08 patch not applied
- OSWorld · 2.0; subset and scoring details unspecified
- OSWorld · 2.0;Opus 5 发布表,未注明评测子集/协议
- OSWorld · 2.1 (September 10, 2026 task/assets release)
- OSWorld · 2.1 offline set; task count and exact task release unspecified in the selected GPT-6.1 Terra article/chart
- OSWorld · 2.1, 82-task offline no-internet subset, partial score
- OSWorld · Qwen3-VL-235B-A22B-Thinking model-card image table; Agent; label as printed; unit and metric definition not stated
- OSWorld · System Card 所称 2.0;2026-09-10 任务与资源版本、108 题、更新 harness
- OSWorld · Version/protocol not stated
- OSWorld · 版本未注明;Claude Opus 4.5 发布材料所列
- OSWorld 2.0 · 版本/单位未注明
- OSWorld-Verified · Partial score; 5 runs; 100-step maximum
- OSWorld-Verified · Verified benchmark; evaluation protocol noted per source
- OSWorld-Verified · 版本/单位未注明 (qg-completion-qwen35-osworld-verified)
- OSWorld-Verified · 版本/单位未注明 (qg-completion-qwen38-27b-osworld-verified)
- OSWorldG · Qwen3-VL-235B-A22B-Thinking model-card image table; Agent; label as printed; unit and metric definition not stated
P
- PaperBench · 10-paper subset; each paper requires <10GB external data
- PaperBench · Subset and metric scale not stated in the comparison table
- Perception Test · Test subset
- PhysicianBench · 100 physician tasks in an electronic health record; pass rate across attempts
- Pile-test · DeepSeek-V3 Technical Report Table 3; Pile-test (BPB); unit is printed as BPB
- PIQA · DeepSeek-V3 Technical Report Table 3; PIQA (EM); unit is not printed
- PLawBench · Task version and metric scale not stated
- PMC-VQA · 版本/单位未注明
- PolyMATH · Qwen3-VL技术报告v2;Table 6; Thinking MoE sampling settings from pages 20-21
- PolyMATH · 版本/单位未注明 (qg-completion-qwen3-235b-2507-polymath)
- PolyMATH · 版本/单位未注明 (qg-completion-qwen35-polymath)
- PolyMATH · 版本/单位未注明 (qs-completion-polymath)
- PostTrainBench · GLM-5.2模型卡;unknown
- PostTrainBench · GLM-5.3模型卡;unknown
- PostTrainBench · v1.1
- PRBench-Finance · Task version and metric scale not stated
- PRBench-Legal · Task version and metric scale not stated
- ProgramBench · 166 个保留的 golden tasks;架构与时限按协议区分
- ProgramBench · GLM-5.2模型卡;unknown
- ProgramBench · Metric Almost@1 as printed in V4.1-Flash card; task version unspecified
- ProgramBench · V4.1-Flash-1017 card does not print metric or task version
- ProgramBench · 版本/指标未注明;Qwen4-Max官方主图/正文;未注明的评测细节保持未知
- ProgramBench (Almost Solved) · GLM-5.3模型卡;Almost Solved
- Protein Design Library Ranking · Prioritization of protein designs for experimental testing
- Protein Design Sequence Generation · Protein sequence generation conditioned on design constraints
- ProteinGym Hard · System-card Hard subset; exact task count not stated
- Protocol Troubleshooting · System-card protocol troubleshooting variant; task count unspecified
- Protocol Understanding · V2; Benchling-created evaluation using real laboratory protocols and applied questions; task count unspecified
- Putnam-200 · Pass@8 with minimal tools; fixed random subset of PutnamBench; exact figure label
Q
R
- RACE-High · DeepSeek-V3 Technical Report Table 3; RACE-High (EM); unit is not printed
- RACE-Middle · DeepSeek-V3 Technical Report Table 3; RACE-Middle (EM); unit is not printed
- Real-World Finance · v2, 294 tasks
- Real-World Finance · v2; 123 real-world financial tasks; one attempt per task in the reported Fable 5.2 run
- RealWorldQA · Qwen3-VL-235B-A22B-Thinking model-card image table; General VQA; label as printed; unit and metric definition not stated
- RealWorldQA · 版本/单位未注明 (qg-completion-qwen35-realworldqa)
- RealWorldQA · 版本/单位未注明 (qg-completion-qwen36-realworldqa)
- RealWorldQA · 版本/单位未注明 (qg-completion-qwen38-27b-realworldqa)
- RealWorldQA · 版本/单位未注明 (qg-completion-qwen38-flash-next-realworldqa)
- Ref-L4 · Test split
- RefCOCO · Average; validation split
- RefCOCO avg · 版本/单位未注明
- RefCOCO-avg · Qwen3-VL-235B-A22B-Thinking model-card image table; 2D/3D Grounding; label as printed; unit and metric definition not stated
- RefCOCO(avg) · 版本/单位未注明
- RefSpatialBench · Qwen3-VL-235B-A22B-Thinking model-card image table; Embedded and Spatial Understanding; label as printed; unit and metric definition not stated
- RefSpatialBench · 版本/单位未注明 (qg-completion-qwen35-refspatialbench)
- RefSpatialBench · 版本/单位未注明 (qg-completion-qwen36-refspatialbench)
- Remote Labor Index · 版本未注明
- RewardBench · DeepSeek-V3 Technical Report Table 8; Average; maj@6; unit not printed
- RewardBench · DeepSeek-V3 Technical Report Table 8; Average; unit not printed
- RiemannBench · 25 private research-mathematics problems; corrected reference answers and grading; four attempts per problem
- RiemannBench · Version not stated in selected source
- RoboDojo Sim · 版本/任务配置未注明
- RoboSpatialHome · Qwen3-VL-235B-A22B-Thinking model-card image table; Embedded and Spatial Understanding; label as printed; unit and metric definition not stated
- RULER 1000k · 1M RULER · DCA · 260 samples/length · unit unspecified
- RULER 1000k · 1M版 · 1000k · YaRN
- RULER 128k · 1M RULER · DCA · 260 samples/length · unit unspecified
- RULER 128k · 1M版 · 128k · YaRN
- RULER 16k · 1M RULER · DCA · 260 samples/length · unit unspecified
- RULER 16k · 1M版 · 16k · YaRN
- RULER 192k · 1M RULER · DCA · 260 samples/length · unit unspecified
- RULER 192k · 1M版 · 192k · YaRN
- RULER 1M average · 1M · 260样本/长度 · YaRN
- RULER 256k · 1M RULER · DCA · 260 samples/length · unit unspecified
- RULER 256k · 1M版 · 256k · YaRN
- RULER 32k · 1M RULER · DCA · 260 samples/length · unit unspecified
- RULER 32k · 1M版 · 32k · YaRN
- RULER 384k · 1M RULER · DCA · 260 samples/length · unit unspecified
- RULER 384k · 1M版 · 384k · YaRN
- RULER 4k · 1M RULER · DCA · 260 samples/length · unit unspecified
- RULER 4k · 1M版 · 4k · YaRN
- RULER 512k · 1M RULER · DCA · 260 samples/length · unit unspecified
- RULER 512k · 1M版 · 512k · YaRN
- RULER 640k · 1M RULER · DCA · 260 samples/length · unit unspecified
- RULER 640k · 1M版 · 640k · YaRN
- RULER 64k · 1M RULER · DCA · 260 samples/length · unit unspecified
- RULER 64k · 1M版 · 64k · YaRN
- RULER 768k · 1M RULER · DCA · 260 samples/length · unit unspecified
- RULER 768k · 1M版 · 768k · YaRN
- RULER 896k · 1M RULER · DCA · 260 samples/length · unit unspecified
- RULER 896k · 1M版 · 896k · YaRN
- RULER 8k · 1M RULER · DCA · 260 samples/length · unit unspecified
- RULER 8k · 1M版 · 8k · YaRN
- RULER 96k · 1M RULER · DCA · 260 samples/length · unit unspecified
- RULER 96k · 1M版 · 96k · YaRN
- RULER Acc avg · 1M RULER · DCA · 260 samples/length · unit unspecified
S
- Scale MultiChallenge · o3-mini grader protocol from GPT-5 developer release
- SciCode · Component of the Artificial Analysis Intelligence Index; version unspecified
- SciCode · Scientific research coding
- ScreenSpot · Qwen3-VL-235B-A22B-Thinking model-card image table; Agent; label as printed; unit and metric definition not stated
- ScreenSpot Pro · Qwen3-VL-235B-A22B-Thinking model-card image table; Agent; label as printed; unit and metric definition not stated
- ScreenSpot Pro · 版本/单位未注明
- ScreenSpot-Pro · 1,581 expert-annotated tasks across 23 professional applications and 3 operating systems
- Seal-0 · 版本/单位未注明
- SIFO · Qwen3-VL-235B-A22B-Thinking model-card image table; Instruction Following; label as printed; unit and metric definition not stated
- SIFO-multiturn · Qwen3-VL-235B-A22B-Thinking model-card image table; Instruction Following; label as printed; unit and metric definition not stated
- Simple-QA-Verified · Anthropic factuality evaluation; Claude Sonnet 4.5 grader; exact dataset revision not stated
- SimpleQA · DeepSeek-R1 paper Table 3; SimpleQA (Correct); unit not printed
- SimpleQA · DeepSeek-V3 Technical Report Table 6; SimpleQA (Correct); unit not printed
- SimpleQA · GLM-4.5 Technical Report Table 6;精确题集修订未注明,单位未注明;指标按§4.2.4原文。
- SimpleQA · GLM-5固定v2报告/模型卡;版本未注明
- SimpleQA · Internal search agent protocol; task version unspecified
- SimpleQA · No search-enabled; report F1
- SimpleQA · Qwen3-VL-235B-A22B-Thinking model-card image table; Knowledge; label as printed; unit and metric definition not stated
- SimpleQA · Version unspecified; Correct metric
- SimpleQA Verified · Kaggle official leaderboard
- SimpleQA-Verified · DeepSeek-V4 report Table 7; SimpleQA-Verified (Pass@1); unit not printed
- SimpleVQA · Qwen3-VL-235B-A22B-Thinking model-card image table; General VQA; label as printed; unit and metric definition not stated
- SimpleVQA · 版本/单位未注明 (qg-completion-qwen35-simplevqa)
- SimpleVQA · 版本/单位未注明 (qg-completion-qwen36-simplevqa)
- SingleCellBench · LatchBio Bioinformatics single-cell RNA sequencing analysis evaluation
- SkillsBench · Public v1.1; 87 tasks; average over three runs; score scale not stated
- SkillsBench Avg5 · 版本/单位未注明
- SLAKE · 版本/单位未注明
- SpatialBench Verified · Verified
- Spider · Coder-Next技术报告;版本/单位未注明
- SpreadsheetBench · Full 912-problem set
- STEM Professor Bench · 64-problem internal research subset; score averaged over up to nine attempts per problem
- Structural Biology · Opus 4.7 announcement figure; dataset, version and question format unspecified
- SUN RGB-D · Qwen3-VL技术报告v2;Table 2; page 19 states Omni3D test, IoU threshold 0.15, confidence 1.0; 3D bbox JSON prompt in Appendix B.5
- SUNRGBD · Qwen3-VL-235B-A22B-Thinking model-card image table; 2D/3D Grounding; label as printed; unit and metric definition not stated
- SUNRGBD · 版本/单位未注明
- SuperGPQA · Qwen3-VL-235B-A22B-Thinking model-card image table; Knowledge; label as printed; unit and metric definition not stated
- SuperGPQA · 版本/单位未注明 (qg-completion-qwen3-235b-2507-supergpqa)
- SuperGPQA · 版本/单位未注明 (qg-completion-qwen35-supergpqa)
- SuperGPQA · 版本/单位未注明 (qg-completion-qwen36-supergpqa)
- SuperGPQA · 版本/单位未注明 (qs-completion-supergpqa)
- SWE Multilingual · DeepSeek-V3.2 report Table 2; SWE Multilingual (Resolved); unit not printed
- SWE Multilingual · DeepSeek-V4 report Table 7; SWE Multilingual (Resolved); unit not printed
- SWE Verified · DeepSeek-R1 paper Table 3; SWE Verified (Resolved); unit not printed
- SWE Verified · DeepSeek-V3.2 report Table 2; SWE Verified (Resolved); unit not printed
- SWE Verified · V3.2-Exp comparison; agent/task version unspecified
- SWE-bench Multilingual · 300 problems, 9 programming languages
- SWE-bench Multilingual · Agent mode; task snapshot unspecified
- SWE-bench Multilingual · GLM-4.7模型卡;Multilingual
- SWE-bench Multilingual · GLM-5固定v2报告/模型卡;题集修订未注明
- SWE-bench Multilingual · Task set version not specified in Qwen model card
- SWE-bench Multilingual · V3.2-Exp comparison; agent/task version unspecified
- SWE-bench Multilingual · 版本/单位未注明 (qg-completion-qwen35-swe-bench-multilingual)
- SWE-bench Multilingual · 版本/单位未注明 (qg-completion-qwen36-swe-bench-multilingual)
- SWE-bench Multilingual · 版本/单位未注明 (qg-completion-qwen38-flash-next-swe-bench-multilingual)
- SWE-Bench Multilingual report score · Coder-Next技术报告;报告协议 · 单位未注明
- SWE-bench Multimodal · Multimodal
- SWE-Bench Pro · GLM-5.1模型卡;Pro
- SWE-bench Pro · GLM-5.2模型卡;Pro
- SWE-bench Pro · Pro
- SWE-bench Pro · Pro; Resolved
- SWE-Bench Pro · Public
- SWE-Bench Pro · Subset/version unspecified in source
- SWE-bench Pro · Version and task-set revision not stated in the chart; source labels Score (%)
- SWE-bench Pro · 版本/单位未注明 (qg-completion-qwen36-swe-bench-pro)
- SWE-bench Pro · 版本/单位未注明 (qg-completion-qwen38-27b-swe-bench-pro)
- SWE-bench Pro · 版本/单位未注明 (qg-completion-qwen38-flash-next-swe-bench-pro)
- SWE-bench Pro · 版本/单位未注明 (qg-completion-qwen38-max-api-swe-bench-pro)
- SWE-Bench Pro report score · Coder-Next技术报告;报告协议 · 单位未注明
- SWE-bench Verified · Agent mode; task snapshot unspecified
- SWE-bench Verified · GLM-4.6模型卡;Verified
- SWE-bench Verified · GLM-4.7模型卡;Verified
- SWE-bench Verified · GLM-5固定v2报告/模型卡;题集修订未注明
- SWE-bench Verified · GPT-5 发布评测固定子集;500 题中排除 23 题,n=477
- SWE-bench Verified · Verified subset (500 problems)
- SWE-bench Verified · Version/task snapshot unspecified in V4 preview table; Resolved
- SWE-bench Verified · 版本/单位未注明 (qg-completion-qwen35-swe-bench-verified)
- SWE-bench Verified · 版本/单位未注明 (qg-completion-qwen36-swe-bench-verified)
- SWE-Lancer · IC Diamond set
- SWE-Lancer · IC SWE set
- SWE-Marathon · GLM-5.2模型卡;unknown
- SWE-Marathon (v1.1) · GLM-5.3模型卡;v1.1
- SWE-MM · 版本/单位未注明
- SWE-rebench · GLM-5固定v2报告/模型卡;January 2026
- SWE-sweep · Version not stated in October 2026 table
- SysBench · GLM-4.5 Technical Report Table 6;精确题集修订未注明,单位未注明;指标按§4.2.4原文。
T
- TAU-Bench · Airline category; release unspecified
- TAU-Bench · Retail category; release unspecified
- TAU1-Airline · 版本/单位未注明 (qg-completion-qwen3-235b-2507-tau1-airline)
- TAU1-Airline · 版本/单位未注明 (qs-completion-tau1-airline)
- TAU1-Retail · 版本/单位未注明 (qg-completion-qwen3-235b-2507-tau1-retail)
- TAU1-Retail · 版本/单位未注明 (qs-completion-tau1-retail)
- TAU2-Airline · Qwen3-VL技术报告v2;Table 6; Thinking MoE sampling settings from pages 20-21
- TAU2-Airline · 版本/单位未注明 (qg-completion-qwen3-235b-2507-tau2-airline)
- TAU2-Airline · 版本/单位未注明 (qs-completion-tau2-airline)
- tau2-bench · Airline category
- tau2-bench · Average across Retail, Airline and Telecom categories
- tau2-Bench · DeepSeek-V3.2 report Table 2; tau2-Bench (Pass@1); category average; unit not printed
- TAU2-Bench · 版本/单位未注明
- TAU2-Retail · Qwen3-VL技术报告v2;Table 6; Thinking MoE sampling settings from pages 20-21
- TAU2-Retail · 版本/单位未注明 (qg-completion-qwen3-235b-2507-tau2-retail)
- TAU2-Retail · 版本/单位未注明 (qs-completion-tau2-retail)
- TAU2-Telecom · Qwen3-VL技术报告v2;Table 6; Thinking MoE sampling settings from pages 20-21
- TAU2-Telecom · 版本/单位未注明 (qg-completion-qwen3-235b-2507-tau2-telecom)
- TAU2-Telecom · 版本/单位未注明 (qs-completion-tau2-telecom)
- TAU3-Bench · 版本/单位未注明
- Terminal Bench 2 · 版本/单位未注明
- Terminal Bench 2.0 · DeepSeek-V3.2 report Table 2; Terminal Bench 2.0 (Acc); unit not printed
- Terminal Bench 2.1 (Terminus) · 版本/单位未注明
- Terminal Bench Hard · GLM-4.7模型卡;Hard
- Terminal-Bench · 2.0; Accuracy as printed in V4 mode table
- Terminal-Bench · 2.0; chart labels metric Score (%) and Terminus-2 json harness
- Terminal-Bench · 2.0; GLM source does not name score/mean-reward metric
- Terminal-Bench · 2.0;代理框架按协议区分
- Terminal-Bench · 2.1
- Terminal-Bench · 2.1 (Pass@1); across agent scaffolds at Max reasoning effort; unit not printed
- Terminal-Bench · 2.1; evaluated with Claude Code at avg@10, 5-hour timeout, max_tokens=131,072; metric scale unspecified
- Terminal-Bench · 2.1; metric not printed in V4-0813/0731 cards
- Terminal-Bench · 2.1; Pass@1 as printed in V4.1-Flash card
- Terminal-Bench · 2.1; source metric reported as score percentage
- Terminal-Bench · 2.1;GLM 官方图仅列数值,具体 score / mean reward 指标未注明
- Terminal-Bench · 3.0
- Terminal-Bench · 3.0; Pass@1 as printed in V4.1-Flash card
- Terminal-Bench · 4.0
- Terminal-Bench · 4.0 · 指标未注明;Qwen4-Max官方主图/正文;未注明的评测细节保持未知
- Terminal-Bench · 4.0; metric not printed in V4.1-Flash-1017 card
- Terminal-Bench · 4.0; Pass@1 as printed in V4.1-Flash card
- Terminal-Bench · GLM-4.7模型卡;2.0
- Terminal-Bench · GLM-5.1模型卡;2.0
- Terminal-Bench · GLM-5.2模型卡;2.1
- Terminal-Bench · GLM-5.3模型卡;3.0
- Terminal-Bench · GLM-5固定v2报告/模型卡;2.0 · 修订题集
- Terminal-Bench · GLM-5固定v2报告/模型卡;2.0 · 原题集
- Terminal-Bench · Task version unspecified in V3.1-Terminus/Exp card
- Terminal-Bench · Terminus 1 framework; task version unspecified
- Terminal-Bench · Version unspecified; official report states a percentage
- Terminal-Bench · 版本未注明;Claude Haiku 4.5 发布材料所列
- Terminal-Bench 2.0 · 版本/单位未注明
- Terminal-Bench 2.0 report score · Coder-Next技术报告;报告协议 · 单位未注明
- Terminal-Bench Science · 0.1; chart-reported resolution rate
- Terminal-Bench-Science · 0.1
- Tool Decathlon · 版本/单位未注明 (qg-completion-qwen35-tool-decathlon)
- Tool Decathlon · 版本/单位未注明 (qg-completion-qwen36-tool-decathlon)
- Tool-Decathlon · DeepSeek-V3.2 report Table 2; Tool-Decathlon (Pass@1); unit not printed
- Tool-Decathlon · GLM-5.1模型卡;unknown
- Tool-Decathlon · GLM-5.2模型卡;unknown
- Tool-Decathlon · GLM-5固定v2报告/模型卡;版本未注明
- Toolathlon · June 2026 release used in selected source
- Toolathlon · Release article does not state benchmark version or scoring protocol (2026-03)
- Toolathlon · Version unspecified; Pass@1
- Toolathlon · Version, subset, and metric unspecified in the 2026-04-23 article
- Toolathlon · Version, subset, and metric unspecified in the 2026-07-09 article
- Toolathlon Verified · Verified subset and task release not stated; Pass@1
- Toolathlon Verified (Pass@1) · 版本/单位未注明
- Toolathlon-Verified · June 2026 finalized release
- Toolathlon-Verified · Version/metric unspecified; reported score
- TreeBench · Version not stated
- TriviaQA · DeepSeek-V3 Technical Report Table 3; TriviaQA (EM); unit is not printed
U
- UniSVG · Qwen3-VL技术报告v2;Table 2; multimodal coding benchmark; specific prompt is not separately shown in Appendix B.10
- USA Mathematical Olympiad · 2026
- USAMO Index · 2025; AMC score + 10 × AIME score
- User-flagged factuality · Refreshed 1,000 de-identified conversations: 229 retained cases and 771 new cases; separate general factuality and reported-issue graders; not comparable with earlier system-card versions
- User-flagged factuality, high-difficulty prompts · Anonymized ChatGPT conversations previously flagged by users for factual errors; exact set version and sample size unspecified
V
- V* · Qwen3-VL技术报告v2;Table 2; thinking target has no + marker, so no tool use is indicated; V* tool prompt documented in Appendix B.9
- V* · 版本/单位未注明 (qg-completion-qwen35-v)
- V* · 版本/单位未注明 (qg-completion-qwen36-v)
- Vals Finance Agent · v2
- Vals Index · v2.2
- Vals Index · Version not stated in September 2026 table
- Vals Science Index · v1
- VATEX · Test subset
- VATEX-ZH · Validation subset
- Vending Bench 2 · GLM-5.1模型卡;2
- Vending-Bench · 2
- Vending-Bench Arena · Not stated in selected source
- Vibe Code Bench · Version not stated in selected source
- Vibe-Eval · Reka image-understanding set
- Video-Index · 版本/指标未注明;Qwen4-Max官方主图/正文;未注明的评测细节保持未知
- Video-MME_without_subtitles · Qwen3-VL技术报告v2;Table 2; thinking-model video multiple-choice prompt in Appendix B.8; max 2,048 frames and 224K visual tokens; 2 fps
- Video-MMMU · Gemini 3 Pro model evaluation; recommended media_resolution=HIGH and temperature=0
- Video-MMMU · Test subset
- Video-MMMU · Test version not stated
- VideoMME · Audio + visual; Long test subset; 0-shot; max 1024 frames at 1 fps
- VideoMME · Full test set; audio + visual + subtitles
- VideoMME · Long subset, with subtitle category
- VideoMME · With subtitles
- VideoMME · Without subtitles
- VideoMME(w sub.) · 版本/单位未注明 (qg-completion-qwen35-videomme-w-sub)
- VideoMME(w sub.) · 版本/单位未注明 (qg-completion-qwen36-videomme-w-sub)
- VideoMME(w/o sub.) · 版本/单位未注明 (qg-completion-qwen35-videomme-w-o-sub)
- VideoMME(w/o sub.) · 版本/单位未注明 (qg-completion-qwen36-videomme-w-o-sub)
- VideoMME(w/o sub) · Qwen3-VL-235B-A22B-Thinking model-card image table; Video; label as printed; unit and metric definition not stated
- VideoMMMU · Qwen3-VL-235B-A22B-Thinking model-card image table; Video; label as printed; unit and metric definition not stated
- VideoMMMU · 版本/单位未注明 (qg-completion-qwen35-videommmu)
- VideoMMMU · 版本/单位未注明 (qg-completion-qwen36-videommmu)
- Vision2Web · 版本/单位未注明 (qg-completion-qwen38-27b-vision2web)
- Vision2Web · 版本/单位未注明 (qg-completion-qwen38-flash-next-vision2web)
- VisualPuzzles · Qwen3-VL技术报告v2;Table 2; Direct prompt is used for the table label; CoT prompt is also documented in Appendix B.1
- VisuLogic · Qwen3-VL-235B-A22B-Thinking model-card image table; STEM&Puzzle; label as printed; unit and metric definition not stated
- VITA-Bench · 版本/单位未注明 (qg-completion-qwen35-vita-bench)
- VITA-Bench · 版本/单位未注明 (qg-completion-qwen36-vita-bench)
- VLMsAreBlind · Qwen3-VL技术报告v2;Table 2; VLMBlind prompt in Appendix B.1; sampling parameters not stated
- VlmsAreBlind · 版本/单位未注明
- VSI-Bench · Qwen3-VL技术报告v2;Table 2; VSI-Bench multiple-choice/open-ended video prompt in Appendix B.6; sampling parameters not stated
W
- WANDR · 500-task wide research benchmark
- WANDR · relaxed score;benchmark version 未注明
- We-Math · Qwen3-VL技术报告v2;Table 2; We-Math prompt in Appendix B.1; sampling parameters not stated
- We-Math · 版本/单位未注明
- WebArena · Official WebArena grader, modified fuzzy_match subgrader
- WebArena · Official WebArena grader; exact task release/subset not stated in card
- WebArena-Verified · Full set
- WebArena-Verified · 版本/单位未注明
- WebDev Arena · Gemini 3 release leaderboard score
- WebDev Arena · Leaderboard version not stated in selected source
- Webquest · MultiQA
- Webquest · SingleQA
- WebVoyager · Protocol not stated
- WebVoyager · SoM protocol
- WeMath · Version not stated
- WideSearch · Task version not stated; average item-F1 over four runs
- WideSearch · 版本/单位未注明 (qg-completion-qwen35-widesearch)
- WideSearch · 版本/单位未注明 (qg-completion-qwen36-widesearch)
- WindowsAA · Qwen3-VL技术报告v2;Table 2; agent benchmark; benchmark-specific prompt is not separately shown in Appendix B.11
- WinoGrande · DeepSeek-V3 Technical Report Table 3; WinoGrande (EM); unit is not printed
- WMT24++ · 版本/单位未注明
- Work360 · 原始得分 · 版本未注明;Qwen4-Max官方主图/正文;未注明的评测细节保持未知
- WorkSpaceBench · Task version and metric scale not stated in the comparison table
- WritingBench · Qwen3-VL-235B-A22B-Thinking model-card image table; Subjective Evaluation; label as printed; unit and metric definition not stated
- WritingBench · 版本/单位未注明 (qg-completion-qwen3-235b-2507-writingbench)
- WritingBench · 版本/单位未注明 (qs-completion-writingbench)
Y
Z
- ZebraLogic · Qwen3-VL-235B-A22B-Thinking model-card image table; Reasoning; label as printed; unit and metric definition not stated
- ZeroBench · Average over 4 runs
- ZEROBench · Qwen3-VL-235B-A22B-Thinking model-card image table; STEM&Puzzle; label as printed; unit and metric definition not stated
- ZeroBench · Sub subset; run count not stated
- ZEROBench · 版本/单位未注明
- ZEROBench_Sub · Qwen3-VL-235B-A22B-Thinking model-card image table; STEM&Puzzle; label as printed; unit and metric definition not stated
- ZEROBench_sub · 版本/单位未注明 (qg-completion-qwen35-zerobench-sub)
- ZEROBench_sub · 版本/单位未注明 (qg-completion-qwen36-zerobench-sub)
- ZeroBench-main · With tools; metric not printed in 1017 card
- ZeroBench-main · With tools; Pass@5 as printed in V4.1-Flash card
专题与其他公开测试
A
B
- BioMysteryBench · Human-difficult subset
- BioMysteryBench · Human-solvable subset
- BioPipelineBench · Anthropic internal unpublished computational-biology evaluation; version and task count unspecified
- BrowseComp-ZH · GLM-5固定v2报告/模型卡;版本未注明
- BrowseComp-zh · Version unspecified; Accuracy
- BrowseComp-zh · 版本/单位未注明
- BrowseCompZh · DeepSeek-V3.2 report Table 2; BrowseCompZh (Pass@1); unit not printed
C
- C-Eval · DeepSeek-R1 paper Table 3; C-Eval (EM); unit not printed
- C-Eval · DeepSeek-V3 Technical Report Table 3; C-Eval (EM); unit is not printed
- C-Eval · DeepSeek-V3 Technical Report Table 6; C-Eval (EM); unit not printed
- C-Eval · Version unspecified; Exact Match
- C-Eval · 版本/单位未注明 (qg-completion-qwen35-c-eval)
- C-Eval · 版本/单位未注明 (qg-completion-qwen36-c-eval)
- C-SimpleQA · DeepSeek-R1 paper Table 3; C-SimpleQA (Correct); unit not printed
- C-SimpleQA · DeepSeek-V3 Technical Report Table 6; C-SimpleQA (Correct); unit not printed
- C-SimpleQA · Version unspecified; Correct metric
- C3 · DeepSeek-V3 Technical Report Table 3; C3 (EM); unit is not printed
- CC-Bench · 52 programming/development tasks across six domains; task release unspecified
- CC-Bench · V1.1; 74 real-world coding tests in Claude Code; exact task release unspecified
- CC-Bench-V2 · GLM-5固定v2报告/模型卡;v2 · Chained Tasks
- CC-Bench-V2 · GLM-5固定v2报告/模型卡;v2 · Engineering
- CC-Bench-V2 · GLM-5固定v2报告/模型卡;v2 · HTML
- CC-Bench-V2 · GLM-5固定v2报告/模型卡;v2 · Next.js
- CC-Bench-V2 · GLM-5固定v2报告/模型卡;v2 · React
- CC-Bench-V2 · GLM-5固定v2报告/模型卡;v2 · Repo Exploration
- CC-Bench-V2 · GLM-5固定v2报告/模型卡;v2 · Svelte
- CC-Bench-V2 · GLM-5固定v2报告/模型卡;v2 · Vue
- CC-Bench-V2 · GLM-5固定v2报告/模型卡;v2 · 整体
- CCPM · DeepSeek-V3 Technical Report Table 3; CCPM (EM); unit is not printed
- Chartography · Fable 5.2 contamination check; 100 original versus redrawn charts with new data/answers; no tools; five runs
- Chinese Creative Writing — Instruction Following · Total DS win share
- Chinese Creative Writing — Writing Quality · Total DS win share
- Chinese Functional Writing (internal) · Total pairwise DS win share
- Chinese SimpleQA · GLM-5固定v2报告/模型卡;版本未注明
- Chinese-SimpleQA · DeepSeek-V4 report Table 7; Chinese-SimpleQA (Pass@1); unit not printed
- CLUEWSC · DeepSeek-R1 paper Table 3; CLUEWSC (EM); unit not printed (ds-completion-r1-cluewsc-em)
- CLUEWSC · DeepSeek-R1 paper Table 3; CLUEWSC (EM); unit not printed (ds-completion-r1-zero-cluewsc-em)
- CLUEWSC · DeepSeek-V3 Technical Report Table 3; CLUEWSC (EM); unit is not printed
- CLUEWSC · DeepSeek-V3 Technical Report Table 6; CLUEWSC (EM); unit not printed
- CMath · DeepSeek-V3 Technical Report Table 3; CMath (EM); unit is not printed
- CMMLU · DeepSeek-V3 Technical Report Table 3; CMMLU (EM); unit is not printed
- CMRC · DeepSeek-V3 Technical Report Table 3; CMRC (EM); unit is not printed
- Coding · GLM-4.5-Air模型卡;Coding Benchmarks: SWE-Bench Verified, Terminal-Bench
- Coding · GLM-4.5模型卡;Coding Benchmarks: SWE-Bench Verified, Terminal-Bench
- Complex Instruction and Multi-Turn Writing · Total pairwise DS win share
- CoWorkBench · In-house long-horizon cowork benchmark; task set and metric scale not stated
- CoWorkBench · 版本/单位未注明 (qg-completion-qwen38-27b-coworkbench)
- CoWorkBench · 版本/单位未注明 (qg-completion-qwen38-flash-next-coworkbench)
- CyberGym · GLM-5.1模型卡;unknown
- CyberGym · GLM-5.3模型卡;unknown
- CyberGym · GLM-5固定v2报告/模型卡;版本未注明
E
F
- Factual error rate on difficult prompts · Internal error-inducing prompts in the GPT-6.1 Terra comparison article; exact dataset version, sample size and user-flagged provenance unspecified.
- FrontierSWE v2 · No-GPU public subset; Mean@5; multi-agent; 20-hour deadline
- FrontierSWE v2 · No-GPU public subset; Mean@5; single-agent; 20-hour deadline
- FullStackBench-en · Coder-Next技术报告;版本/单位未注明
- FullStackBench-zh · Coder-Next技术报告;版本/单位未注明
G
H
I
- IF-Badcase · GLM-5固定v2报告/模型卡;版本未注明
- INCLUDE · Qwen3-VL-235B-A22B-Thinking model-card image table; Multilingual; label as printed; unit and metric definition not stated
- INCLUDE · 版本/单位未注明 (qg-completion-qwen3-235b-2507-include)
- INCLUDE · 版本/单位未注明 (qg-completion-qwen35-include)
- INCLUDE · 版本/单位未注明 (qs-completion-include)
- Internal API instruction-following evaluation · Hard subset; exact set/version unspecified
L
- LLM Performance Evaluation composite · GLM-4.5-Air模型卡;12 benchmarks: MMLU-Pro, AIME 24, MATH-500, SciCode, GPQA, HLE, LCB (2407-2501), SWE-Bench Verified, Terminal-Bench, TAU-Bench, BFCL V3, BrowseComp
- LLM Performance Evaluation composite · GLM-4.5模型卡;12 benchmarks: MMLU-Pro, AIME 24, MATH-500, SciCode, GPQA, HLE, LCB (2407-2501), SWE-Bench Verified, Terminal-Bench, TAU-Bench, BFCL V3, BrowseComp
M
- MENT-SNS · GLM-5固定v2报告/模型卡;版本未注明
- MMBench · V1.1 Chinese
- MMBench_EN_V1.1_dev · Qwen3-VL-235B-A22B-Thinking model-card image table; General VQA; label as printed; unit and metric definition not stated
- MMBench-CN · Qwen3-VL技术报告v2;Table 2; MMBench prompt in Appendix B.2; sampling parameters not stated
- MMBench-EN · Qwen3-VL技术报告v2;Table 2; MMBench prompt in Appendix B.2; sampling parameters not stated
- MMBenchEN-DEV-v1.1 · 版本/单位未注明 (qg-completion-qwen35-mmbenchen-dev-v1-1)
- MMBenchEN-DEV-v1.1 · 版本/单位未注明 (qg-completion-qwen36-mmbenchen-dev-v1-1)
- MMLU Language (0-shot) · OpenAI GPT-5 system card, 2025-08-07; MMLU test set translated by professional human translators; chain-of-thought prompting. Section 3.10 says 13 languages, but Table 10 lists 14 language rows.
- MMLU-ProX · Qwen3-VL-235B-A22B-Thinking model-card image table; Multilingual; label as printed; unit and metric definition not stated
- MMLU-ProX · 版本/单位未注明 (qg-completion-qwen3-235b-2507-mmlu-prox)
- MMLU-ProX · 版本/单位未注明 (qg-completion-qwen35-mmlu-prox)
- MMLU-ProX · 版本/单位未注明 (qs-completion-mmlu-prox)
- MMMLU · DeepSeek-V3 Technical Report Table 3; MMMLU non-English (EM); unit is not printed
- MultiIF · Qwen3-VL-235B-A22B-Thinking model-card image table; Multilingual; label as printed; unit and metric definition not stated
- MultiIF · 版本/单位未注明 (qs-completion-multiif)
O
- OCRBench_v2_en · Qwen3-VL技术报告v2;Table 2; OCRBench_v2 prompt in Appendix B.4; sampling parameters not stated
- OCRBench_v2_zh · Qwen3-VL技术报告v2;Table 2; OCRBench_v2 prompt in Appendix B.4; sampling parameters not stated
- OCRBenchV2_(en/ch) · Qwen3-VL-235B-A22B-Thinking model-card image table; Text Recognition and Chart/Document Understanding; label as printed; unit and metric definition not stated
- OmniDocBench_en · Qwen3-VL技术报告v2;Table 2; OmniDocBench Markdown-conversion prompt in Appendix B.4; metric direction is not stated in Table 2 caption
- OmniDocBench_zh · Qwen3-VL技术报告v2;Table 2; OmniDocBench Markdown-conversion prompt in Appendix B.4; metric direction is not stated in Table 2 caption
- Organic Chemistry · Anthropic internal unpublished life-science evaluation; version unspecified
P
Q
- QwenClawBench · 版本/单位未注明
- QwenQoderBench · In-house Qoder UX benchmark; avg@5; 6-hour timeout; task release not stated
- QwenReactBench · In-house bilingual seven-category React benchmark; auto-render and multimodal judge
- QwenSVGBench · In-house bilingual SVG benchmark; auto-render and multimodal judge; task release not stated
- QwenSWEBench · In-house coding benchmark; avg@3; 8-hour timeout; task release not stated
- QwenSWEBench · 版本/单位未注明
- QwenWebBench · 版本/单位未注明
R
- R&D Coding Benchmark (internal) · 30-task internal engineering evaluation; Pass Rate (%)
- Real-World Finance · Anthropic internal evaluation; about 50 analyst-workflow tasks; version unspecified
- Reasoning · GLM-4.5-Air模型卡;Reasoning Benchmarks: MMLU-Pro, AIME 24, MATH 500, SciCode, GPQA, HLE, LCB (2407-2501)
- Reasoning · GLM-4.5模型卡;Reasoning Benchmarks: MMLU-Pro, AIME 24, MATH 500, SciCode, GPQA, HLE, LCB (2407-2501)
- RecreationBench · 版本/单位未注明 (qg-completion-qwen38-27b-recreationbench)
- RecreationBench · 版本/单位未注明 (qg-completion-qwen38-flash-next-recreationbench)
- RewardBench · DeepSeek-V3 Technical Report Table 8; Chat-Hard; maj@6; unit not printed
- RewardBench · DeepSeek-V3 Technical Report Table 8; Chat-Hard; unit not printed
- RewardBench · DeepSeek-V3 Technical Report Table 8; Chat; maj@6; unit not printed
- RewardBench · DeepSeek-V3 Technical Report Table 8; Chat; unit not printed
- RewardBench · DeepSeek-V3 Technical Report Table 8; Reasoning; maj@6; unit not printed
- RewardBench · DeepSeek-V3 Technical Report Table 8; Reasoning; unit not printed
- RewardBench · DeepSeek-V3 Technical Report Table 8; Safety; maj@6; unit not printed
- RewardBench · DeepSeek-V3 Technical Report Table 8; Safety; unit not printed
S
T
- Template Following generalization average · Coder-Next技术报告;版本/单位未注明
- Template Following generalization scaffold-1 · Coder-Next技术报告;版本/单位未注明
- Template Following generalization scaffold-2 · Coder-Next技术报告;版本/单位未注明
- Template Following generalization scaffold-3 · Coder-Next技术报告;版本/单位未注明
- Template Following generalization scaffold-4 · Coder-Next技术报告;版本/单位未注明
- Template Following generalization scaffold-5 · Coder-Next技术报告;版本/单位未注明
- Toolcall-Badcase · GLM-5固定v2报告/模型卡;版本未注明