FinRiskAtlas - Evidence-Grounded Processing - Institution Matching: leaderboard
FinRiskAtlas fixed-evidence review operation: deciding whether two institution descriptions, possibly from different records or languages, denote the same institution, 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: Not forecast. 33 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 93.6 | #90 |
| 2 | Qwen3-Next-80B | 90.8 | #321 |
| 3 | GPT-5.6 Terra | 90.4 | #40 |
| 4 | Qwen 3 32B | 90 | #449 |
| 5 | GPT-5.6 Sol | 90 | #22 |
| 6 | Gemini 2.5 Flash | 89.2 | #255 |
| 7 | DeepSeek V4 Flash | 88 | #155 |
| 8 | Qwen 3 235B A22B | 88 | #323 |
| 9 | GLM-5 | 87.6 | #149 |
| 10 | Kimi K2.5 | 87.2 | #152 |
| 11 | DeepSeek V4 Pro | 86.8 | #112 |
| 12 | Seed 2.0 Mini | 86.8 | #279 |
| 13 | GLM-5.2 | 85.6 | #80 |
| 14 | GPT-5.6 Luna | 85.6 | #79 |
| 15 | Kimi K3 | 84.8 | #35 |
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-institution-matching · How It Works · Data refreshed daily, snapshot 2026-11-05.