🧪 Test?View on arXiv
Finding the Right Evidence: Factor-Guided Coarse-to-Fine Reasoning for Long Videos
Not provided in the abstract
evidence acquisitionlong-video reasoningmultimodalcontrastive learning
2608.26355
Builder Relevance
1h ago80%
Abstract
The paper presents PACE, a framework that improves long-video question answering by enhancing evidence acquisition through a two-stage process.
Reality Card
Core Claim
PACE achieves 42.6% accuracy on long-video question answering, outperforming previous methods by focusing on option-discriminative evidence.
Method / Result
PACE recovers 66.9% of the annotated cues, indicating improved evidence recovery.
Limitations
The paper does not specify the authors, which may affect reproducibility and transparency.
Paper to code
Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.
No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.