LexKairos - Bounded Event Retrieval: leaderboard
The Bounded Event Retrieval sub-task of LexKairos (Tsinghua University, 2026): retrieve the case events that fall inside a time window. Accuracy in percent over 594 items in the vanilla setting (direct answer); higher is better.
Metric: Accuracy (%; 594 items: retrieve the case events inside a time window, single- and multi-label; vanilla setting, the model answers directly without extra reasoning steps (thinking disabled where the model allows it); temperature 0, 8,000 output tokens). Source: arxiv.org. Saturation forecast: Around February 2027. 8 models tracked. Updated 30 Sep 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | DeepSeek R1 | 85.69 | #478 | |
| 2 | GPT-5.4 | Non-reasoning | 84.68 | #128 |
| 3 | Gemini 3 Flash | Minimal | 74.24 | #163 |
| 4 | DeepSeek V4 Flash | Non-reasoning | 57.24 | #259 |
| 5 | LegalOne-8B | Non-reasoning | 38.21 | #4746 |
| 6 | Qwen 3.5 9B | Non-reasoning | 36.36 | #814 |
| 7 | Llama 3.1 8B | 35.35 | #5129 | |
| 8 | DISC-LawLLM-7B | 34.85 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=lexkairos-bounded-event-retrieval · How It Works · Data refreshed daily, snapshot 2026-11-16.