FCMBench-Video (English) - Evidence-Grounded Selection: leaderboard

FCMBench-Video (2026) English-language handheld document videos from financial credit workflows, task evidence-grounded selection: exact-match accuracy (%) on single-choice questions answered from the right document in a multi-document video. Zero-shot; higher is better.

Metric: Evidence-Grounded Selection: exact-match accuracy (%) on single-choice questions answered from the right document in a multi-document video, on FCMBench-Video's en-US subset (795 composed videos of 8 English document types, 5,362 instances); zero-shot with structured outputs; each model's native video serving path (2 FPS sampling for most open models, 0.5 FPS for Ovis2.5, raw video upload for the API models); higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 8 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Qwen 3.5 27B86.85#375
2Gemini 3 Pro (Preview)76.99#119
3Doubao-Seed-1.6-vision55.96#361
4Ovis2.5-9B42.2#1225
5Qwen 3 VL 32B Instruct41.9#606
6Qwen3 Omni 30B A3B Instruct38.53#930
7Qwen 3 VL 8B Instruct33.03#1029
8InternVL3-8B32.72#1782

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=fcmbench-video-english-evidence-grounded-selection · How It Works · Data refreshed daily, snapshot 2026-11-21.