LegalCiteTrust: leaderboard
LegalCiteTrust evaluates Chinese long-form legal research reports along three axes: coverage of the requested research structure, the legal evidence supplied, and whether each cited statute or case exists, is described faithfully and supports the claim it is cited for. The final score is coverage plus support multiplied by that citation trust score, so it runs from zero to two and rewards well-covered reports only when their authorities hold up.
Metric: Final score (0 to 2; per report (Coverage + Support) x Trust, each dimension in 0-1, macro-averaged over the 72 Chinese legal research tasks; Coverage of the rubric issues, Support by statute and case evidence, Trust as the mean product of Existence, Fidelity and Applicability of each cited law or case; each system in its own usage mode, direct LLMs without tools). Source: arxiv.org. Saturation forecast: Around February 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 0.98 | #167 |
| 2 | Qwen 3.6 Plus | 0.96 | #222 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=legalcitetrust · How It Works · Data refreshed daily, snapshot 2026-11-13.