OmniToM - Belief Labeling - Order: leaderboard
OmniToM (2026) evaluates theory of mind through explicit belief modeling on 895 short social stories from ToMBench, with gold belief propositions built by a human-calibrated annotation pipeline. This board is Stage 2 labeling accuracy for the belief order (0 for world facts, 1 to 3 for recursive beliefs) dimension of the given belief propositions; higher is better.
Metric: Belief-labeling accuracy (%): exact-match accuracy of the predicted schema label on the given gold belief propositions, macro-averaged over stories, for the belief order (0 for world facts, 1 to 3 for recursive beliefs) dimension, on the 895 OmniToM stories from seven ToMBench categories (22,343 gold belief propositions), zero-shot TELeR level 3 prompts; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 9 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Mistral Large 2 (Jul) | 97.25 | #1144 |
| 2 | Gemma 3 27B (IT) | 96.56 | #1324 |
| 3 | Qwen 3 32B | 96.42 | #972 |
| 4 | Gemini 2.5 Flash | 95.56 | #463 |
| 5 | GPT-5 | 95.18 | #178 |
| 6 | Mistral Small 3 | 95.13 | #1796 |
| 7 | Llama 3.3 70B Instruct | 92.74 | #1385 |
| 8 | Qwen 3 8B | 73.38 | #1873 |
| 9 | Llama 3.1 8B Instruct | 71.9 | #4600 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=omnitom-belief-labeling-order · How It Works · Data refreshed daily, snapshot 2026-11-21.