AgentOCR: Reimagining Agent History via Optical Self-Compression

TL;DR

AgentOCR leverages visual tokens and self-compression to significantly reduce token usage in agentic systems, maintaining high performance while improving efficiency and scalability.

Contribution

This paper introduces a novel visual token-based history representation with segment optical caching and adaptive self-compression for scalable, efficient agent systems.

Findings

Preserves over 95% of text-based agent performance

Reduces token consumption by more than 50%

Achieves a 20x speedup in rendering

Abstract

Recent advances in large language models (LLMs) enable agentic systems trained with reinforcement learning (RL) over multi-turn interaction trajectories, but practical deployment is bottlenecked by rapidly growing textual histories that inflate token budgets and memory usage. We introduce AgentOCR, a framework that exploits the superior information density of visual tokens by representing the accumulated observation-action history as a compact rendered image. To make multi-turn rollouts scalable, AgentOCR proposes segment optical caching. By decomposing history into hashable segments and maintaining a visual cache, this mechanism eliminates redundant re-rendering. Beyond fixed rendering, AgentOCR introduces agentic self-compression, where the agent actively emits a compression rate and is trained with compression-aware reward to adaptively balance task success and token efficiency. We conduct extensive experiments on challenging agentic benchmarks, ALFWorld and search-based QA. Remarkably, results demonstrate that AgentOCR preserves over 95\% of text-based agent performance while substantially reducing token consumption (>50\%), yielding consistent token and memory efficiency. Our further analysis validates a 20x rendering speedup from segment optical caching and the effective strategic balancing of self-compression.