AI Paper Atlas

Research map / Video / Video-language understanding

Chain-of-Frames: Advancing Video Understanding in Multimodal LLMs via Frame-Aware Reasoning

Sara Ghazanfari, Francesco Croce, Nicolas Flammarion, Prashanth Krishnamurthy, Farshad Khorrami, Siddharth Garg

CVPR 2026 · Video › Video-language understanding · approach: Video LLMs with fine-grained understanding · 39 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
fine-tune on chain-of-frames traces instead of standard promptingInternVL3-8B · Average of 5 benchmarksaccuracy: 67.0 -> 72.1Improved
fine-tune on chain-of-frames traces (ablation)InternVL2.5-4B · VSI-Benchaccuracy: 31.8 -> 36.9Improved
add CoT prompting without fine-tuning (ablation)InternVL2.5-4B · Video-MMEaccuracy: 54.9 -> 54.7No clear effect
add CoT prompting without fine-tuning (ablation)InternVL2.5-4B · VidHalaccuracy: 74.0 -> 77.0Improved
fine-tune on QA pairs only (no reasoning) (ablation)InternVL2.5-4B · VidHalaccuracy: 74.0 -> 64.1Hurt
fine-tune on QA pairs only (no reasoning) (ablation)InternVL2.5-4B · MVBenchaccuracy: 70.8 -> 73.4Improved
fine-tune on CoT traces without frame references (ablation)InternVL2.5-4B · EventHallusionaccuracy: 62.5 -> 53.1Hurt
add explicit frame references to CoT traces (ablation)InternVL2.5-4B · EventHallusionaccuracy: 53.1 -> 71.2Improved
add explicit frame references to CoT traces (ablation)InternVL2.5-4B · VSI-Benchaccuracy: 34.3 -> 36.9Improved
train on synthetic CLEVRER-based CoF data vs real VideoEspresso data (equal size) (ablation)InternVL2.5-4B · five video benchmarksaccuracyImproved
combine real and synthetic CoF data (ablation)InternVL2.5-4B · five video benchmarksaccuracyMixed
CoF fine-tuning vs multi-stage video CoT (M-LLM)InternVL2.5-4B · NExT-QAaccuracy: 75.3 -> 79.6Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy