Search papers, labs, and topics across Lattice.
The paper identifies that Video-LLMs struggle with physical reasoning due to "Semantic Prior Dominance," where internal narrative scripts override visual perception. To address this, they introduce the Programmatic Adversarial Curriculum (PACC), a high-fidelity adversarial video dataset designed to decouple visual artifacts from logical errors. They also propose the Visual-Anchored Reasoning Chain (VARC) to encourage grounding judgments in low-level visual facts, achieving significant improvements in physical reasoning via LoRA fine-tuning without architectural changes.
Video-LLMs aren't failing at perception, they're being tricked by their own assumptions, but a new dataset and reasoning chain can fix it.
While Video Large Language Models (Video-LLMs) excel in general understanding, they exhibit systematic deficits in fine-grained physical reasoning. Existing interventions not only suffer from limited generalization but fundamentally conflate generative artifacts with genuine physical fallacies. Furthermore, we find that models fail systematically not only in anti-physics anomalies but also in counter-intuitive scenarios where visual facts contradict statistical expectations. Accordingly, we propose the Unified Attribution Theory: this dual failure stems not from perception deficiency, but from Semantic Prior Dominance -- the reasoning mechanism is deeply hijacked by internal narrative scripts. To address this, we construct the Programmatic Adversarial Curriculum (PACC), the first high-fidelity adversarial video dataset synthesized based on physical laws, thoroughly decoupling visual artifacts from logical errors. Concurrently, we design the Visual-Anchored Reasoning Chain (VARC) to force models to explicitly ground their judgments in low-level visual facts prior to logical adjudication. Experiments demonstrate that without invasive architectural modifications, standard LoRA fine-tuning with the PACC curriculum effectively neutralizes prior interference in state-of-the-art (SOTA) models, yielding a substantial leap in physical reasoning capabilities.
