Reproduce “AdaAlloc: Adaptive Visual Token Allocation for Long-Video Question Answering” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Joungbin An, Kristen Grauman
AdaAlloc allocates a fixed visual-token budget between sparse low-resolution video-wide frames and denser high-resolution views of selected moments, using question-conditioned planning and temporal grounding followed by refinement. With Qwen3-VL-8B-Instruct, a matched 32K final-answering budget yields 56.6% accuracy on LVBench, versus 46.1% for uniform sampling. The evaluation covers four benchmarks and 7–8B backbones.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.