HELM TORR - Turl Col Type - Robustness: leaderboard
HELM ToRR (release v1.0.0) robustness on Turl Col Type (annotating table column types (TURL); 41 instances): per instance, 1 minus the spread between the best and worst Micro F1 across 35 prompt configurations (7 table serializations x original, row-shuffled, column-shuffled, transposed and padded tables), averaged. 0-100, higher is more consistent; a model that fails every configuration also scores high.
Metric: Robustness (0-100). Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.5 Sonnet (20241022) | 62.22 | #302 |
| 2 | Gemini 1.5 Flash (002) | 59.29 | #587 |
| 3 | GPT-4o (2024-11-20) | 57.76 | #392 |
| 4 | DeepSeek V3 | 56.08 | #329 |
| 5 | Qwen 2 72B Instruct | 51.58 | #579 |
| 6 | Claude 3.5 Haiku (20241022) | 51.26 | #590 |
| 7 | Llama 3.1 70B Instruct | 50.94 | #567 |
| 8 | GPT-4o Mini (2024-07-18) | 50.64 | #568 |
| 9 | Gemini 1.5 Pro (002) | 50.6 | #368 |
| 10 | Llama 3.1 405B Instruct | 49.82 | #470 |
| 11 | Mixtral 8x22B Instruct | 45.36 | #738 |
| 12 | Mistral 7B Instruct (v0.3) | 27.36 | #1255 |
| 13 | Llama 3.1 8B Instruct | 18.87 | #1017 |
| 14 | Mixtral 8x7B Instruct | 17.08 | #1085 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-torr-turl-col-type-robustness · How It Works · Data refreshed daily, snapshot 2026-11-05.