AI Paper Atlas

Research map / Video / Video-language understanding

Fostering Video Reasoning via Next-Event Prediction

Haonan Wang, Hongfu Liu, Xiangyan Liu, Chao Du, Kenji Kawaguchi, Ye Wang, Tianyu Pang

ICLR 2026 · Video › Video-language understanding · approach: Video question answering · 13 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
train on next-event prediction instead of captioning/MCQA/OEQAQwen2.5-VL-7B-Instruct · Temporal benchmarks averageaccuracy: 49.7 -> 53.5Improved
train on NEPQwen2.5-VL-7B-Instruct · General benchmarks averageaccuracy: 60.3 -> 60.9No clear effect
use next-event (deductive) instead of video QA or previous-event prediction (ablation)Qwen2.5-VL-7B-Instruct · FutureBenchaccuracy: 58.8 (QA) / 55.1 (PEP) -> 61.3Improved
use previous-event prediction instead of NEP (ablation)Qwen2.5-VL-7B-Instruct · SeedBench-R1accuracy: 39.5 -> 31.2Hurt
use Mix tuning instead of SFT (ablation)Qwen2.5-VL-7B-Instruct · Temporal benchmarks averageaccuracy: 52.6 -> 53.3Improved
use Distill instead of SFT (ablation)Qwen2.5-VL-7B-Instruct · General benchmarks averageaccuracy: 59.7 -> 61.2Improved
use critique fine-tuning instead of SFT (ablation)Qwen2.5-VL-7B-Instruct · FutureBenchaccuracy: 61.3 -> 55.8Hurt
increase NEP training data beyond 5K samples (ablation)SFT/CFT/Distill/Mix · General and temporal benchmarksaccuracyNo clear effect
train with GRPO instead of SFT on NEP-style MCQA (ablation)Qwen2.5-VL-7B-Instruct · FutureBench 1-Hopaccuracy: 67.6 -> 83.8Improved
train with GRPO instead of SFT on NEP-style MCQA (ablation)Qwen2.5-VL-7B-Instruct · FutureBench 3-Hop (OOD)accuracy: 57.7 -> 62.7Improved
train with GRPO instead of SFT (ablation)Qwen2.5-VL-7B-Instruct · General benchmarks averageaccuracy: 59.7 -> 58.2Hurt
train with captioning on full videoQwen2.5-VL-7B-Instruct · TempCompassaccuracy: 73.8 -> 72.2Hurt

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