BeliefShift (RAG) - Belief Revision Accuracy: leaderboard
Metric: Belief Revision Accuracy (0-1 times 100): how closely the model's tracked belief state follows the annotated change after evidence-driven sessions, normalized by the size of the true shift, over BeliefShift trajectories, with the top-k earlier sessions retrieved by a dense index; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o | 83 | #333 |
| 2 | Gemini 1.5 Pro | 79 | #442 |
| 3 | Claude 3.5 Sonnet | 74 | #337 |
| 4 | Mistral Large | 65 | #693 |
| 5 | Gemini 1.5 Flash | 60 | #619 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=beliefshift-rag-belief-revision-accuracy · How It Works · Data refreshed daily, snapshot 2026-10-11.