FinRiskAtlas - Evidence-Grounded Processing - Information Extraction: leaderboard
FinRiskAtlas fixed-evidence review operation: extracting an ordered list of requested evidence fields from a source record, scored by normalized field-level comparison (FinRiskAtlas, arXiv 2608.25325v1, Appendix Table 9). Percent; higher is better. Reported for 33 archived model configurations under zero-shot direct-answer inference.
Metric: Field-Level Score (%). Source: arxiv.org. Saturation forecast: Around February 2027. 33 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.6 Luna | 85.97 | #79 |
| 2 | GPT-5.6 Sol | 85.07 | #22 |
| 3 | Seed 2.0 Lite | 84.48 | #161 |
| 4 | Kimi K2.5 | 84.28 | #152 |
| 5 | GPT-5.4 | 83.98 | #83 |
| 6 | GPT-5.4 Mini | 83.78 | #228 |
| 7 | GPT-5.6 Terra | 83.48 | #40 |
| 8 | Qwen 3.7 Max | 82.79 | #85 |
| 9 | GLM-5.2 | 81.69 | #80 |
| 10 | Nemotron 3 Ultra | 81.49 | #245 |
| 11 | GLM-5 | 81.39 | #149 |
| 12 | Qwen 3 235B A22B | 81.25 | #323 |
| 13 | DeepSeek V3 (0324) | 79.12 | #349 |
| 14 | Qwen 3 14B | 76.82 | #542 |
| 15 | Claude Opus 4.7 | 76.62 | #51 |
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-information-extraction · How It Works · Data refreshed daily, snapshot 2026-11-05.