More Thinking, Less Seeing? Assessing Amplified Hallucination in Multimodal Reasoning Models
This addresses the trade-off between reasoning ability and perceptual fidelity in multimodal AI, which is incremental but important for improving model reliability in tasks like multimodal math reasoning.
The study tackled the problem of increased hallucination in multimodal large language models as reasoning chains lengthen, finding that larger models typically achieve a better balance between reasoning and perception, influenced more by training data types and domains than volume.
Test-time compute has empowered multimodal large language models to generate extended reasoning chains, yielding strong performance on tasks such as multimodal math reasoning. However, this improved reasoning ability often comes with increased hallucination: as generations become longer, models tend to drift away from image-grounded content and rely more heavily on language priors. Attention analysis shows that longer reasoning chains lead to reduced focus on visual inputs, which contributes to hallucination. To systematically study this phenomenon, we introduce RH-AUC, a metric that quantifies how a model's perception accuracy changes with reasoning length, allowing us to evaluate whether the model preserves visual grounding during reasoning. We also release RH-Bench, a diagnostic benchmark that spans a variety of multimodal tasks, designed to assess the trade-off between reasoning ability and hallucination. Our analysis reveals that (i) larger models typically achieve a better balance between reasoning and perception, and (ii) this balance is influenced more by the types and domains of training data than by its overall volume. These findings underscore the importance of evaluation frameworks that jointly consider both reasoning quality and perceptual fidelity.