JudgeSense - Coherence: leaderboard
JudgeSense (2026) measures how often an LLM judge gives the same verdict when one judging request is worded two equivalent ways, at temperature 0, on items from human-labelled corpora. Agreement on the exact 1 to 5 coherence rating of SummEval summaries across the two wordings, from 0 to 1; higher is better.
Metric: Judge Sensitivity Score: share of items on which the judge returns the same verdict under two reworded instruction templates that ask for the same judgement, over the pairs it answered (refusals and malformed outputs left out); temperature 0 (claude-opus-4-7 at the provider default), 1,024 output tokens, on 250 SummEval summaries rated 1 to 5 for coherence, agreement on the exact rating; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 25 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Mistral Medium (2604) | 0.892 | #679 |
| 2 | Qwen 2.5 72B Instruct | 0.86 | #1117 |
| 3 | Gemma 4 31B (IT) | 0.848 | #350 |
| 4 | Gemini 3.7 Flash | 0.84 | #41 |
| 5 | Kimi K3 | 0.832 | #61 |
| 6 | Claude Sonnet 4.5 | 0.828 | #261 |
| 7 | magistral-small-latest (JudgeSense checkpoint unspecified) | 0.82 | |
| 8 | mistral-small-latest (JudgeSense checkpoint unspecified) | 0.804 | |
| 9 | Claude Haiku 4.5 (20251001) | 0.792 | #543 |
| 10 | Qwen 3.7 Flash (2026-07-15) | 0.788 | |
| 11 | Llama 3.3 70B Instruct | 0.772 | #1385 |
| 12 | Llama 4 Maverick Instruct FP8 | 0.764 | #1023 |
| 13 | Qwen 3.8 27B | 0.76 | #221 |
| 14 | Qwen 3.6 35B A3B | 0.76 | #481 |
| 15 | Claude Opus 4.7 | 0.736 | #85 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=judgesense-coherence · How It Works · Data refreshed daily, snapshot 2026-11-21.