FCMBench V1.1 - Validity Checking: leaderboard
Evaluates visual credit-document understanding on FCMBench V1.1: 26 document types, 5,198 images and 13,806 VQA pairs. Exact-match key/value sets are scored with F1; task scores macro-average fine-grained subtasks. Numeric calculation permits differences below 2. The paper prefers instruct or lowest available reasoning effort. The two-stage DeepSeek-OCR plus V3.2 system stays distinct from either model alone. These are paper V1.1 scores, not the differing repository leaderboard. This series reports FCMBench V1.1 - Validity Checking using F1 (%).
Metric: F1 (%). Source: arxiv.org. Saturation forecast: Around February 2027. 28 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 87.27 | #82 |
| 2 | Gemini 3 Flash | 73.06 | #103 |
| 3 | Qwen 3.5 397B A17B | 71.87 | #154 |
| 4 | Claude Opus 4.5 | 71.21 | #87 |
| 5 | Kimi K2.5 | 66.65 | #152 |
| 6 | GLM-4.5V | 61.16 | #358 |
| 7 | Qwen 3 VL 32B Instruct | 60.8 | #290 |
| 8 | GLM-4.6V | 60.6 | #330 |
| 9 | GPT-5.2 | 60.05 | #115 |
| 10 | Llama 4 Maverick | 59.78 | #486 |
| 11 | GPT-5.1 | 59.16 | #145 |
| 12 | Qwen 3.5 122B A10B | 58.04 | #184 |
| 13 | Qwen 3 VL 235B A22B Instruct | 57.37 | #280 |
| 14 | Qwen 3 VL 30B A3B Instruct | 55.36 | #390 |
| 15 | Qwen 3.5 35B A3B | 54.38 | #268 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=fcmbench-v1-1-validity-checking · How It Works · Data refreshed daily, snapshot 2026-11-05.