Search papers, labs, and topics across Lattice.
CADER introduces a novel framework for long-video understanding that dynamically adjusts its reasoning process based on answer confidence, allowing for more efficient use of tool-assisted reasoning. By employing a two-stage approach, CADER first assesses confidence through global reasoning and exits early for high-confidence cases, while uncertain examples undergo a more intensive evidence localization process. Experiments demonstrate that CADER not only enhances reasoning performance across various VideoQA benchmarks but also competes effectively with specialized tool-augmented systems, all without requiring additional training.
CADER redefines long-video reasoning by enabling systems to adaptively allocate resources based on confidence, significantly improving efficiency and accuracy.
Long-video understanding increasingly relies on large vision-language models and tool-augmented reasoning, but most systems apply the same inference procedure to every example regardless of difficulty. This uniform strategy invokes unnecessary tool-assisted processing for easy questions and provides limited control when difficult questions require fine-grained temporal evidence. We propose CADER (Confidence-Aware Dynamic Evidence Reasoning), a training-free framework for adaptive and reliable long-video reasoning. CADER first performs global reasoning over uniformly sampled frames and estimates answer confidence with a logit-margin signal, allowing high-confidence examples to exit early. For uncertain examples, CADER activates a second-stage tool-augmented loop that combines temporal cropping, lightweight semantic verification, and Relevance-Guided Resampling to progressively localize question-relevant evidence. This design treats tool use as a sample-level decision: a single global pass handles easy cases, while additional reasoning is reserved for examples where uncertainty suggests that more evidence is needed. Experiments on multiple VideoQA benchmarks show that CADER improves long-video reasoning while bypassing Stage~2 for high-confidence samples. Moreover, when applied to a backbone trained only with tool-free chain-of-thought supervision, CADER achieves competitive performance against specialized tool-augmented frameworks, suggesting a practical inference-time route for adaptive long-video reasoning.
Tsinghua AI