AI Paper Atlas

Research map / Video / Video-language understanding

EgoVITA: Learning to Plan and Verify for Egocentric Video Reasoning

Yogesh Kulkarni, Pooyan Fazli

ECCV 2026 · Video › Video-language understanding · approach: Egocentric video understanding · 7 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
add the full EgoVITA plan-verify RL trainingQwen2.5-VL-7B · EgoBlindaccuracy: 29.7 -> 37.4Improved
add dense rewards (ACMG and confidence) on top of format+answer GRPOQwen2.5-VL-7B · EgoBlindaccuracy: 35.6 -> 37.4Improved
apply SFT only on egocentric dataQwen2.5-VL-7B · MVBenchaccuracy: 65.2 -> 63.2Hurt
add RL (EgoVITA) over SFTQwen2.5-VL-7B · Video-MMEaccuracy: 60.1 -> 64.6Improved
large-scale egocentric training (EgoThinker, ~5M samples)Qwen2-VL · DocVQAaccuracy: 79.5 -> 69.0Hurt
add teacher-guided warm-up in confidence reward (ablation)EgoVITA · EgoBlindaccuracy: 36.0 -> 37.4Improved
add ACMG reward alone to format+answer GRPO (ablation)Qwen2.5-VL-7B · EOC-Benchaccuracy: 42.2 -> 43.1Improved
add confidence reward alone to format+answer GRPO (ablation)Qwen2.5-VL-7B · EOC-Benchaccuracy: 42.2 -> 43.5Improved
combine ACMG and confidence rewards (ablation)Qwen2.5-VL-7B · EOC-Benchaccuracy: 43.5 -> 44.1Improved
add egocentric plan stage (ablation)EgoVITA · EgoPlanaccuracy: 29.6 -> 32.9Improved
add exocentric verification stage (ablation)EgoVITA · Video-MMEaccuracy: 61.2 -> 64.6Improved
use anticipatory future-frame grounding instead of present-frame (ablation)EgoVITA · EgoBlindaccuracy: 33.8 -> 37.4Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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