HELM TORR - Tablebench Fact Checking: leaderboard
The "Tablebench Fact Checking" 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: ROUGE-L. Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o (2024-11-20) | 72.56 | #378 |
| 2 | DeepSeek V3 | 71.07 | #323 |
| 3 | Gemini 1.5 Pro (002) | 70.13 | #350 |
| 4 | Gemini 1.5 Flash (002) | 69.92 | #585 |
| 5 | Claude 3.5 Sonnet (20241022) | 69.61 | #291 |
| 6 | Mixtral 8x22B Instruct | 67.77 | #733 |
| 7 | Qwen 2 72B Instruct | 67.61 | #579 |
| 8 | GPT-4o Mini (2024-07-18) | 65.73 | #561 |
| 9 | Llama 3.1 405B Instruct | 65.22 | #461 |
| 10 | Llama 3.1 70B Instruct | 63.38 | #566 |
| 11 | Claude 3.5 Haiku (20241022) | 62.57 | #591 |
| 12 | Mixtral 8x7B Instruct | 58.8 | #1172 |
| 13 | Mistral 7B Instruct (v0.3) | 54.94 | #1276 |
| 14 | Llama 3.1 8B Instruct | 51.55 | #1040 |
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-tablebench-fact-checking · How It Works · Data refreshed daily, snapshot 2026-10-23.