HELM TORR - Turl Col Type: leaderboard
The "Turl Col Type" subtask of HELM TORR. Stanford HELM's TORR project (Table-Oriented Reasoning and Retrieval): question answering and reasoning over tables and semi-structured records, including financial filings, scored on program and answer accuracy.
Metric: Micro F1. Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.5 Sonnet (20241022) | 67.48 | #293 |
| 2 | DeepSeek V3 | 65.59 | #325 |
| 3 | Gemini 1.5 Pro (002) | 61.86 | #352 |
| 4 | Llama 3.1 405B Instruct | 61.54 | #464 |
| 5 | Qwen 2 72B Instruct | 60.44 | #581 |
| 6 | Llama 3.1 70B Instruct | 59.62 | #567 |
| 7 | GPT-4o (2024-11-20) | 59.34 | #379 |
| 8 | Gemini 1.5 Flash (002) | 57.18 | #584 |
| 9 | Claude 3.5 Haiku (20241022) | 55.25 | #591 |
| 10 | Mixtral 8x22B Instruct | 54.32 | #735 |
| 11 | GPT-4o Mini (2024-07-18) | 54.13 | #560 |
| 12 | Mixtral 8x7B Instruct | 35.17 | #1171 |
| 13 | Mistral 7B Instruct (v0.3) | 30.44 | #1272 |
| 14 | Llama 3.1 8B Instruct | 18 | #1042 |
No result here: #4 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-torr-turl-col-type · How It Works · Data refreshed daily, snapshot 2026-10-24.