MM-THEBench: Do Reasoning MLLMs Think Reasonably?
This work addresses the issue of measuring hallucinations in reasoning MLLMs for researchers and developers, but it is incremental as it builds on existing benchmark efforts by focusing on internal thinking processes.
The paper tackles the problem of hallucinations in multimodal large language models (MLLMs) during reasoning processes by introducing MM-THEBench, a benchmark that assesses intermediate chain-of-thought hallucinations, revealing insights into how thinking affects hallucination and reasoning capability across tasks.
Recent advances in multimodal large language models (MLLMs) mark a shift from non-thinking models to post-trained reasoning models capable of solving complex problems through thinking. However, whether such thinking mitigates hallucinations in multimodal perception and reasoning remains unclear. Self-reflective reasoning enhances robustness but introduces additional hallucinations, and subtle perceptual errors still result in incorrect or coincidentally correct answers. Existing benchmarks primarily focus on models before the emergence of reasoning MLLMs, neglecting the internal thinking process and failing to measure the hallucinations that occur during thinking. To address these challenges, we introduce MM-THEBench, a comprehensive benchmark for assessing hallucinations of intermediate CoTs in reasoning MLLMs. MM-THEBench features a fine-grained taxonomy grounded in cognitive dimensions, diverse data with verified reasoning annotations, and a multi-level automated evaluation framework. Extensive experiments on mainstream reasoning MLLMs reveal insights into how thinking affects hallucination and reasoning capability in various multimodal tasks.