Search papers, labs, and topics across Lattice.
The paper addresses the problem of hallucination in Large Vision-Language Models (LVLMs) by identifying and exploiting a Generative-Discriminative Gap, where LVLMs are better at verifying details than generating them. They introduce OSCAR, an Online Self-Calibration framework that uses Monte Carlo Tree Search with a Dual-Granularity Reward Mechanism to generate preference data for Direct Preference Optimization. Experiments show OSCAR reduces hallucinations and improves general multimodal capabilities, achieving state-of-the-art performance on hallucination benchmarks.
LVLMs are better at spotting their own mistakes than generating correct answers in the first place, and this self-awareness can be exploited to reduce hallucinations.
Large Vision-Language Models (LVLMs) often suffer from hallucinations, generating descriptions that include visual details absent from the input image. Recent preference alignment methods typically rely on supervision distilled from stronger models such as GPT. However, this offline paradigm introduces a Supervision-Perception Mismatch: the student model is forced to align with fine-grained details beyond its perceptual capacity, learning to guess rather than to see. To obtain reliable self-supervision for online learning, we identify a Generative-Discriminative Gap within LVLMs, where models exhibit higher accuracy on discriminative verification than open-ended generation. Leveraging this capability, we propose \textbf{O}nline \textbf{S}elf-\textbf{CA}lib\textbf{R}ation (OSCAR), a framework that integrates Monte Carlo Tree Search with a Dual-Granularity Reward Mechanism to construct preference data and iteratively refines the model via Direct Preference Optimization. Extensive experiments demonstrate that OSCAR achieves state-of-the-art performance on hallucination benchmarks while improving general multimodal capabilities.
