AI Paper Atlas

Research map / Video / Video-language understanding

TSPO: Temporal Sampling Policy Optimization for Long-form Video Language Understanding

Canhui Tang, Zifan Han, Hongbo Sun, Sanping Zhou, Xuchong Zhang, Xin Wei, Ye Yuan, Huayu Zhang, Jinglin Xu, Hao Sun

AAAI 2026 · Video › Video-language understanding · approach: Video LLMs with fine-grained understanding · 24 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 TSPO temporal agent to LLaVA-Video-7B (64 frames)LLaVA-Video-7B · MLVU Devaccuracy: 70.3 -> 76.3Improved
add TSPO temporal agent to LLaVA-Video-7B (64 frames)LLaVA-Video-7B · Video-MME (Average)accuracy: 64.4 -> 65.5Improved
add TSPO to LLaVA-Video-7BLLaVA-Video-7B · LVBenchaccuracy: 40.2 -> 45.3Improved
transfer TSPO agent trained on LLaVA-Video to Qwen2.5-VL without retrainingQwen2.5-VL-7B · MLVU Devaccuracy: 65.1 -> 74.3Improved
transfer TSPO agent to LLaVA-Video-72BLLaVA-Video-72B · MLVU Devaccuracy: 74.4 -> 77.3Improved
train on Comprehensive Temporal data (ablation)TSPO (LLaVA-Video-7B) · LongVideoBench / Video-MMEaccuracy: 58.9 -> 62.8 / 64.4 -> 65.5Improved
train on Needle-in-a-Haystack data only (ablation)TSPO (LLaVA-Video-7B) · LongVideoBench / Video-MMEaccuracy: 58.9 -> 63.4 / 64.4 -> 64.6Mixed
combine both data styles (ablation)TSPO (LLaVA-Video-7B) · LongVideoBench / Video-MMEaccuracy: 63.8 / 65.0 (accuracy reward only)Improved
add temporal locating reward to accuracy reward (ablation)TSPO (LLaVA-Video-7B) · LongVideoBench / Video-MMEaccuracy: 63.8 / 65.0 -> 63.9 / 65.5Improved
use RL (TSPO) instead of end-to-end SFT with Gumbel-Softmax (ablation)temporal agent · LongVideoBench / Video-MMEaccuracy: 62.8 / 64.8 -> 63.9 / 65.5Improved
use TSPO selector instead of CoS 13B selectorLLaVA-Video-7B · Frame selection time (128 candidates)frame selection time: 28.4 -> 1.2Improved
reduce frames from 64 to 32 with TSPO (ablation)LLaVA-Video + TSPO · LongVideoBench / Video-MMEaccuracy: 60.6 / 65.3 -> 59.6 / 64.8Mixed

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