Research map / Vision-language and multimodal / Robot action and navigation
MemOCR: Layout-Aware Visual Memory for Efficient Long-Horizon Reasoning
Yaorui Shi, Shugui Liu, Yu Yang, Wenyu Mao, Yuxin Chen, Qi GU, Hui Su, Xunliang Cai, Xiang Wang, An Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add T_augM (compressed-memory QA) training objective (ablation) | MemOCR · Avg, 10K, B=16 | accuracy (%): 41.2 -> 62.2 | Improved |
| add T_augM training objective (ablation) | MemOCR · Avg, 10K, B=64 | accuracy (%): 48.5 -> 67.3 | Improved |
| add T_augM training objective (ablation) | MemOCR · Avg, 10K, B=1024 | accuracy (%): 61.3 -> 74.6 | Improved |
| add T_augQ (detail question) objective on top of T_std (ablation) | MemOCR · Avg, 10K, B=16 | accuracy (%): 32.9 -> 41.2 | Improved |
| add T_augQ objective on top of T_std (ablation) | MemOCR · Avg, 10K, B=1024 | accuracy (%): 62.2 -> 61.3 | No clear effect |
| add RL training with all objectives versus none (ablation) | MemOCR · Avg, 10K, B=16 | accuracy (%): 37.5 -> 62.2 | Improved |
| add T_std alone versus untrained model (ablation) | MemOCR · Avg, 10K, B=64 | accuracy (%): 55.9 -> 41.1 | Improved |
| replace text truncation with summarization for MemAgent | MemAgent · Avg, 10K, B=16 | accuracy (%): 31.6 -> 58.5 | Improved |
| use visual memory instead of textual summary memory | MemOCR vs MemAgent · Avg, 10K, B=16 | accuracy (%): 31.6 -> 62.2 | Improved |
| give text baselines 4x more tokens than MemOCR | MemAgent (B=64) vs MemOCR (B=16) · Avg, 10K | accuracy (%): 50.7 -> 62.2 | Improved |
| lower memory budget from 1024 to 16 on TriviaQA | MemOCR · TriviaQA 10K | accuracy (%): 79.6 -> 80.8 | No clear effect |
| use MemOCR instead of unlimited-storage RAG memory HyMem | MemOCR (B=1024) · Avg, 10K | accuracy (%): 56.4 -> 74.6 | Improved |
Closest papers by meaning
- ViLoMem: Agentic Learner with Grow-and-Refine Multimodal Semantic Memory (CVPR 2026 · similarity 0.90)
- Memory-T1: Reinforcement Learning for Temporal Reasoning in Multi-session Agents (ICLR 2026 · similarity 0.90)
- Memory is Reconstructed, Not Retrieved: Graph Memory for LLM Agents (ICML 2026 · similarity 0.90)
- Navigating Massive Visual Context in Retrieval-Augmented Generation via Multimodal Memory Graph (ICML 2026 · similarity 0.90)
- Seeing, Listening, Remembering, and Reasoning: A Multimodal Agent with Long-Term Memory (ICLR 2026 · similarity 0.89)
- Evaluating Memory in LLM Agents via Incremental Multi-Turn Interactions (ICLR 2026 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
