LongJud-Bench - Position Consistency (Second Instance): leaderboard

Metric: Position Consistency Score (%; Claude Sonnet 4.6 judge, 0-10 per trial phase on consistency of the second-instance trial statements with the appeal or response, normalized). Source: arxiv.org. Saturation forecast: Around December 2026. 6 models tracked.

Top models

#ModelScore
1Kimi K2.565
2GPT-5.264
3Qwen 3.5 Plus64
4DeepSeek V4 Flash63
5GLM-4.761
6Qwen 3.5 Flash55

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