AI Paper Atlas

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

ICML 2026 · Vision-language and multimodal › Robot action and navigation · approach: Embodied multimodal LLM agents · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
add T_augM (compressed-memory QA) training objective (ablation)MemOCR · Avg, 10K, B=16accuracy (%): 41.2 -> 62.2Improved
add T_augM training objective (ablation)MemOCR · Avg, 10K, B=64accuracy (%): 48.5 -> 67.3Improved
add T_augM training objective (ablation)MemOCR · Avg, 10K, B=1024accuracy (%): 61.3 -> 74.6Improved
add T_augQ (detail question) objective on top of T_std (ablation)MemOCR · Avg, 10K, B=16accuracy (%): 32.9 -> 41.2Improved
add T_augQ objective on top of T_std (ablation)MemOCR · Avg, 10K, B=1024accuracy (%): 62.2 -> 61.3No clear effect
add RL training with all objectives versus none (ablation)MemOCR · Avg, 10K, B=16accuracy (%): 37.5 -> 62.2Improved
add T_std alone versus untrained model (ablation)MemOCR · Avg, 10K, B=64accuracy (%): 55.9 -> 41.1Improved
replace text truncation with summarization for MemAgentMemAgent · Avg, 10K, B=16accuracy (%): 31.6 -> 58.5Improved
use visual memory instead of textual summary memoryMemOCR vs MemAgent · Avg, 10K, B=16accuracy (%): 31.6 -> 62.2Improved
give text baselines 4x more tokens than MemOCRMemAgent (B=64) vs MemOCR (B=16) · Avg, 10Kaccuracy (%): 50.7 -> 62.2Improved
lower memory budget from 1024 to 16 on TriviaQAMemOCR · TriviaQA 10Kaccuracy (%): 79.6 -> 80.8No clear effect
use MemOCR instead of unlimited-storage RAG memory HyMemMemOCR (B=1024) · Avg, 10Kaccuracy (%): 56.4 -> 74.6Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy