Get Started
Home
Topics
Search
Library
Research questionHow can long-video QA organize multimodal memory to preserve temporal and cross-modal grounding under limited context?Long-video QA systems often store captions, frames, transcripts, summaries, and graph facts as separate fragments. At answer time, models must reconstruct which modalities refer to the same event and when it occurred, despite limited context.
AI
AI Memory
Evaluation & Benchmarks
Inference Optimization
Information Retrieval
Multimodal Models
Retrieval-Augmented Generation
Latest papersRecent research connected to this question, newest first.EM^2Mem: Event-Centric Multimodal Memory for Large Language ModelsThe source presents EM^2Mem, an event-centric multimodal memory framework that links heterogeneous records to event anchors and retains temporal context, relations, semantic facts, and provenance. Its reported evidence comes from three long-video QA benchmarks, including accuracy, strict event-level Top-5 evidence recall, per-query latency, and inference-token measurements.research paper · Sep 1, 2026
Related questions
How can long-video agents choose evidence-acquisition strategies for focused, broad-coverage, or contrastive questions?How can long-term conversational QA agents retrieve and reason over temporally dispersed dialogue history?How can multimodal models maintain useful visual memory for causal streaming video reasoning under fixed memory?How can continual VideoQA learn new tasks without forgetting earlier ones or accumulating task-specific prompts?