Text2DistBench - Distribution Estimation: leaderboard
Text2DistBench (2026) tests distributional reading comprehension: given viewer comments on newly released movies and songs, a model reports how opinions are distributed over topics and sentiments rather than single facts. This board scores estimated distributions as 1 minus total variation distance, from 0 to 1, averaged over marginal, conditional and joint questions; higher is better.
Metric: 1 minus total variation distance (0-1) between the predicted and empirical opinion distribution of viewer comments, averaged over marginal, conditional and joint questions, 20 movie and music entities released July-October 2025; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 6 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.1 | 0.881 | #240 |
| 2 | Grok 4 Fast | 0.868 | #447 |
| 3 | Gemini 2.5 Pro | 0.859 | #276 |
| 4 | Claude Sonnet 4.5 | 0.851 | #261 |
| 5 | Llama 3.3 70B | 0.812 | #1514 |
| 6 | Qwen 3 32B | 0.689 | #972 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=text2distbench-distribution-estimation · How It Works · Data refreshed daily, snapshot 2026-11-21.