ResearchResearch paperLarge Language Models · Interpretability1 source · Oct 7, 2026

Less from More: Reinforcing Sparse Video Reasoning from Dense References

We propose SAVER, a dense-to-sparse post-training framework that uses dense video views as training-time references for sparse-frame inference.

Video-language models commonly assume that more temporal observations lead to more reliable reasoning.
We question this assumption and argue that the key challenge is not merely processing more video frames efficiently, but learning to reason reliably under limited temporal evidence.
During reinforcement post-training, paired dense and sparse views are optimized with grounding rewards and a reliability-gated reference reward, encouraging sparse view predictions to preserve task-relevant temporal evidence.
These results show that temporal grounding can serve as an effective evidence-localization proxy for learning sparse video reasoning that transfers to broader video understanding tasks.

[1]Less from More: Reinforcing Sparse Video Reasoning from Dense References
arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 08:51 PM
We propose SAVER, a dense-to-sparse post-training framework that uses dense video views as training-time references for sparse-frame inference.
Video-language models commonly assume that more temporal observations lead to more reliable reasoning.

Extractive summary: sentences quoted from the sources.

Before this