AI Paper Atlas

Research map / Video / Video-language understanding

Video-in-the-Loop: Span-Grounded Long Video QA with Interleaved Reasoning

Chendong Wang, Donglin Bai, Yifan Yang, Xiao Jin, Anlan Zhang, Rui Wang, Shiqi Jiang, Yuqing Yang, Hao Wu, Qi Dai, Chong Luo, Ting Cao and 2 more

ICML 2026 · Video › Video-language understanding · approach: Video question answering · 0 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 frame-level textual timestamp injection (ablation)single-stage SFT (Qwen2.5-VL) · LVBenchaccuracy (%): 40.2 -> 44.3Improved
add frame-level textual timestamp injection (ablation)single-stage SFT (Qwen2.5-VL) · LongVideoBenchaccuracy (%): 61.3 -> 62.1Improved
use Stage-2 answering prompt on the full video (ablation)SFT + TI · LVBenchaccuracy (%): 44.3 -> 45.4Improved
add learned localization and interleaved GRPO with coupled QA+TVG rewards (ablation)ViTL · LVBenchaccuracy (%): 45.4 -> 47.9Improved
add learned localization and interleaved GRPO (ablation)ViTL · LongVideoBenchaccuracy (%): 62.5 -> 63.5Improved
replace CLIP top-k frame selection with learned parametric selection (ablation)ViTL · LongVideoBenchaccuracy (%): 58.1 -> 63.3Improved
use predicted spans instead of random zoom (ablation)ViTL · LongVideoBenchaccuracy (%): 55.2 -> 63.3Improved
use ground-truth spans instead of predicted spans (oracle) (ablation)ViTL · LongVideoBenchaccuracy (%): 63.3 -> 70.5Improved
zoom to high fidelity instead of reusing low-fps frames (Refine-only) (ablation)ViTL · LongVideoBenchaccuracy (%): 56.2 -> 63.3Improved
add iterative reasoning without zoom (Refine-only) (ablation)single-stage Qwen2.5-VL · LongVideoBenchaccuracy (%): 56.0 -> 56.2No clear effect
allow multiple disjoint spans instead of a single span (ablation)ViTL · LVBenchaccuracy (%): 45.1 -> 47.4Improved
add two-stage skim-zoom pipelineViTL · latency, 128 frameslatency (s): 6.24 -> 7.32Hurt

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