FinRiskAtlas - Evidence-Grounded Processing - Person Matching: leaderboard
FinRiskAtlas fixed-evidence review operation: deciding whether two partially overlapping person records denote the same natural person, scored by accuracy (FinRiskAtlas, arXiv 2608.25325v1, Appendix Table 9). Percent; higher is better. Reported for 33 archived model configurations under zero-shot direct-answer inference.
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around February 2027. 33 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K3 | 87 | #35 |
| 2 | GPT-5.6 Sol | 86 | #22 |
| 3 | GPT-5.4 Mini | 85.8 | #228 |
| 4 | Gemini 3 Flash (Preview) | 85.8 | #90 |
| 5 | GPT-5.6 Luna | 85.8 | #79 |
| 6 | GPT-5.6 Terra | 85.8 | #40 |
| 7 | Gemini 2.5 Flash | 85.6 | #255 |
| 8 | GPT-5.4 | 85.6 | #83 |
| 9 | Nemotron 3 Ultra | 85.6 | #245 |
| 10 | Qwen 3 14B | 85.4 | #542 |
| 11 | DeepSeek V4 Pro | 85.2 | #112 |
| 12 | Qwen 3.7 Max | 84.4 | #85 |
| 13 | Qwen 3 32B | 84.36 | #449 |
| 14 | GLM-5 | 84.2 | #149 |
| 15 | Qwen 3 4B | 84 | #834 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=finriskatlas-evidence-grounded-processing-person-matching · How It Works · Data refreshed daily, snapshot 2026-11-05.