M$^3$Exam: Benchmarking Multimodal Memory for Realistic User-Agent Interactions
For developers of language agents handling multimodal data, this benchmark and method address the lack of realistic evaluation and efficiency in multimodal memory systems.
M^3Exam benchmarks multimodal memory in realistic user-agent interactions, revealing persistent gaps in cross-modal grounding and reasoning. The proposed M^3Proctor method improves accuracy by 13% while reducing index-construction time and retrieved tokens by over 70%.
Language agents are increasingly deployed over accumulating multimodal information, yet existing benchmarks assume a human-human form with sparse visuals and straightforward content, evaluating neither reasoning over authentic multimodal file interaction nor the interpretation of concealed user information. We therefore introduce M$^3$Exam, a query-centric multimodal conversational memory benchmark built on realistic user-agent interaction, with multi-dimensional evaluation spanning cross-modal grounding and implicit information inference. Benchmarking MLLMs and memory systems reveals persistent gaps in cross-modal grounding, cross session reasoning, and the efficiency cost of accumulating multimodal context. We further propose M$^3$Proctor, a multimodal memory method that detects query modality bias and consumes raw visual sources only on demand, improving accuracy by 13% while cutting index-construction time and retrieved tokens by over 70%.