DDR-Bench - 10-K - Item-Averaged Trajectory-Wise: leaderboard
Evaluates autonomous database exploration with SQL/Python tools on MIMIC, GLOBEM and 10-K. GPT-5 mini checks coverage of 2,058 checklist items across 291 task entities. Message-wise and trajectory-wise insights use sample- or item-averaging. Paper v1 scores are preserved; current repository defaults include a 100-turn safety cap. Released financial-trajectory labels resolve eight Qwen instruct/checkpoint identities; ambiguous Qwen3-30B-A3B and Kimi K2 snapshots remain explicit source-local rows. This series reports DDR-Bench - 10-K - Item-Averaged Trajectory-Wise using Checklist Coverage Accuracy (%).
Metric: Checklist Coverage Accuracy (%). Source: arxiv.org. Saturation forecast: Around February 2027. 22 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 61.25 | #148 |
| 2 | GPT-5.1 | 44.76 | #145 |
| 3 | GPT-5.2 | 41.22 | #115 |
| 4 | DeepSeek V3.2 | 38.16 | #210 |
| 5 | GPT-5 Mini | 36.77 | #193 |
| 6 | GLM-4.6 | 36.16 | #265 |
| 7 | Qwen 3 Next 80B A3B Instruct | 31.1 | #381 |
| 8 | MiniMax-M2 | 26.82 | #327 |
| 9 | Gemini 3 Flash | 21.08 | #103 |
| 10 | Qwen 3 4B 2507 Instruct | 19.91 | #768 |
| 11 | Gemini 2.5 Flash | 16.49 | #255 |
| 12 | Gemini 2.5 Pro | 16.14 | #159 |
| 13 | Qwen 2.5 72B Instruct | 14.02 | #469 |
| 14 | Qwen 2.5 14B Instruct | 10.12 | #654 |
| 15 | Qwen2.5-14B-Instruct-1M | 9.78 | #581 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=ddr-bench-10-k-item-averaged-trajectory-wise · How It Works · Data refreshed daily, snapshot 2026-11-05.