HELM TORR - Wikitq - Robustness: leaderboard
HELM ToRR (release v1.0.0) robustness on WikiTableQuestions (question answering over Wikipedia tables; 60 instances): per instance, 1 minus the spread between the best and worst F1 Strings 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) | 80.19 | #302 |
| 2 | DeepSeek V3 | 78.21 | #329 |
| 3 | Llama 3.1 405B Instruct | 76.09 | #470 |
| 4 | Llama 3.1 70B Instruct | 75.87 | #567 |
| 5 | Qwen 2 72B Instruct | 75.34 | #579 |
| 6 | GPT-4o (2024-11-20) | 75.33 | #392 |
| 7 | GPT-4o Mini (2024-07-18) | 73.3 | #568 |
| 8 | Gemini 1.5 Flash (002) | 72.33 | #587 |
| 9 | Gemini 1.5 Pro (002) | 69.01 | #368 |
| 10 | Llama 3.1 8B Instruct | 66.83 | #1017 |
| 11 | Claude 3.5 Haiku (20241022) | 60.78 | #590 |
| 12 | Mistral 7B Instruct (v0.3) | 39.85 | #1255 |
| 13 | Mixtral 8x22B Instruct | 39.1 | #738 |
| 14 | Mixtral 8x7B Instruct | 28.81 | #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-wikitq-robustness · How It Works · Data refreshed daily, snapshot 2026-11-05.