LongJud-Bench - Fact Marshalling (First Instance): leaderboard

Metric: Fact Marshalling Score (%; Claude Sonnet 4.6 judge, 0-10 rubric on the facts and reasons in the first-instance complaint or defense, normalized). Source: arxiv.org. Saturation forecast: Around December 2026. 6 models tracked.

Top models

#ModelScore
1Kimi K2.568
2DeepSeek V4 Flash67
3GLM-4.767
4Qwen 3.5 Plus66
5Qwen 3.5 Flash62
6GPT-5.260

Interactive version: theaggregate.ai/benchmark?slug=longjud-bench-fact-marshalling-first-instance · How It Works · Data refreshed daily, snapshot 2026-09-25.