Research
Benchmarks
Compare state-of-the-art methods across the tasks and datasets used to measure AI progress.
309,131 benchmarksPage 1813 of 12,366
Benchmarks
| Task | Dataset | Domain | Trend | Results | Last update |
|---|---|---|---|---|---|
| Code Generation | HumanEval pass@1 (test) | 18 | Jul 7, 2026 | ||
| Fine-grained VQA | V* | 18 | Jun 1, 2026 | ||
| Long-context reasoning | LongReason | 18 | Jun 1, 2026 | ||
| Mathematical Reasoning | AIME | 18 | Jun 1, 2026 | ||
| Mathematical Reasoning | MATH | 18 | Jun 1, 2026 | ||
| Mathematical Reasoning | GSM8K | 18 | Jun 1, 2026 | ||
| Out-of-Distribution Detection | ImageNet-C | 18 | Jun 17, 2026 | ||
| Out-of-Distribution Detection | ImageNet V2 | 18 | Jun 17, 2026 | ||
| Math & Code Reasoning | ARC Easy | 18 | Jun 18, 2026 | ||
| Long-context language understanding | RULER 16K | 18 | Jun 1, 2026 | ||
| Long-context language understanding | RULER 16K 1.0 (test) | 18 | Jun 1, 2026 | ||
| Long Video Event Prediction | MILES | 18 | Jun 1, 2026 | ||
| Long Video Event Prediction | CMD | 18 | Jun 1, 2026 | ||
| Absolute Free Energy Estimation | Ising Model | 18 | Jun 1, 2026 | ||
| 3D Object Detection | nuScenes Clean | 18 | Jun 1, 2026 | ||
| Safety Risk Assessment | Held-out Set | 18 | Jun 1, 2026 | ||
| LLM Unlearning | ArXiv unlearning Llama2-7B | 18 | Jun 1, 2026 | ||
| Jailbreak Attack | AdvCUA | 18 | Jun 1, 2026 | ||
| Jailbreak Attack | AgentHarm | 18 | Jun 1, 2026 | ||
| Aggregate Performance Evaluation | MMLU, GSM, HellaSwag, TruthfulQA, ARC-C, CodeX | 18 | Jun 1, 2026 | ||
| Agentic re-identification | 27 transcripts Gemini-3-Flash attacker | 18 | Jun 1, 2026 | ||
| Agentic re-identification | 27 transcripts GPT-5.4-mini attacker | 18 | Jun 1, 2026 | ||
| Agentic re-identification | 27 transcripts GPT-5.1 attacker | 18 | Jun 1, 2026 | ||
| Mission Planning | BCCM Failure Scenarios Part-A 1.0 | 18 | Jun 1, 2026 | ||
| Mathematical Reasoning | HMMT-26 | 18 | Jun 1, 2026 |