From Label Priors to Task Evidence: Efficient Long-Video Frame Selection via Bayesian GRPO
Multimodal Large Language Models (MLLMs) face significant scalability challenges in long-video understanding due to quadratic attention complexity and context window constraints.
ProofPaper ↗
Key points
- Existing retrieval-based approaches typically rely on supervised learning from static keyframe annotations.
- To bridge this gap, we propose a probabilistic framework that reformulates frame selection as latent-variable inference.
- Then, we treat a frozen MLLM as a task environment and use its answer likelihood as dense task evidence to refine the policy via Bayesian Group Relative Policy Optimization (GRPO).
- Extensive experiments on Video-MME, MLVU, and LongVideoBench demonstrate that our selector achieves competitive accuracy with MLLM-based scorers while retaining the inference efficiency of lightweight methods.
Sources (1)
- [1]From Label Priors to Task Evidence: Efficient Long-Video Frame Selection via Bayesian GRPOarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 18, 05:17 PM
Multimodal Large Language Models (MLLMs) face significant scalability challenges in long-video understanding due to quadratic attention complexity and context window constraints.
Existing retrieval-based approaches typically rely on supervised learning from static keyframe annotations.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 18, 2026Google Open-Sources Harness2, a Recursive Agent Harnessing Framework
- Oct 16, 2026GeoReason: Bridging Logical Reasoning and Spatial Fidelity in Remote Sensing Segmentation
- Oct 16, 2026Xiaomi-Lens-0 Technical Report
- Oct 16, 2026Welcome Beam: a 501B open model you can own, customize, and scale
- Oct 16, 2026DLite: History-free Speculative Decoding for Agentic AI
- Oct 16, 2026Tree-Aware Posterior Distillation for Long-Horizon Agentic Reinforcement Learning