Search papers, labs, and topics across Lattice.
The paper introduces MuRGAt, a new benchmark for evaluating fact-level multimodal attribution in complex reasoning scenarios involving video, audio, and other modalities. MuRGAt requires models to generate answers with explicit reasoning and precise citations, specifying modality and temporal segments for each claim. The authors also develop an automatic evaluation framework that correlates well with human judgment, revealing that current MLLMs struggle with accurate citation even when reasoning is correct, and that enforcing grounding can degrade accuracy.
Even state-of-the-art multimodal LLMs struggle to accurately cite their sources when reasoning across video, audio, and text, often hallucinating citations despite generating correct answers.
Multimodal large language models (MLLMs) are increasingly used for real-world tasks involving multi-step reasoning and long-form generation, where reliability requires grounding model outputs in heterogeneous input sources and verifying individual factual claims. However, existing multimodal grounding benchmarks and evaluation methods focus on simplified, observation-based scenarios or limited modalities and fail to assess attribution in complex multimodal reasoning. We introduce MuRGAt (Multimodal Reasoning with Grounded Attribution), a benchmark for evaluating fact-level multimodal attribution in settings that require reasoning beyond direct observation. Given inputs spanning video, audio, and other modalities, MuRGAt requires models to generate answers with explicit reasoning and precise citations, where each citation specifies both modality and temporal segments. To enable reliable assessment, we introduce an automatic evaluation framework that strongly correlates with human judgments. Benchmarking with human and automated scores reveals that even strong MLLMs frequently hallucinate citations despite correct reasoning. Moreover, we observe a key trade-off: increasing reasoning depth or enforcing structured grounding often degrades accuracy, highlighting a significant gap between internal reasoning and verifiable attribution.
