FinRiskAtlas - Applied Review - Disputed-Issue Generation: leaderboard
Metric: Rubric Score (%; DeepSeek-V4-Flash judge). Source: arxiv.org. 32 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | DeepSeek V4 Pro | 77.12 |
| 2 | Nemotron 3 Super | 76.67 |
| 3 | Kimi K3 | 74.84 |
| 4 | Gemini 3 Flash (Preview) | 74.69 |
| 5 | Claude Opus 4.7 | 74.06 |
| 6 | Gemini 2.5 Flash | 74.05 |
| 7 | Seed 2.0 Lite | 73.35 |
| 8 | Seed 2.0 Mini | 72.73 |
| 9 | GLM-5 | 72.49 |
| 10 | Nemotron 3 Ultra | 71.58 |
| 11 | Qwen 3.7 Max | 70.58 |
| 12 | Qwen 3 4B | 68.83 |
| 13 | GLM-5.2 | 68.37 |
| 14 | Qwen3-Next-80B | 67.51 |
| 15 | Ling-2.6-1T | 63.3 |
Interactive version: theaggregate.ai/benchmark?slug=finriskatlas-applied-review-disputed-issue-generation · How It Works · Data refreshed daily, snapshot 2026-09-19.