The hidden state immediately before each agent action is studied and it is found that compression and recall needs are already encoded in the model's internal representations, which substantially reduces context consumption while maintaining competitive task performance.
Abstract
Long horizon language model agents continuously accumulate interaction history, increasing computational cost while making relevant information harder to preserve and reuse. Existing context management methods mainly focus on how to compress or retrieve history, but largely leave open whether the model itself already represents the need for these memory operations before they occur. We study the hidden state immediately before each agent action and find that compression and recall needs are already encoded in the model's internal representations. These signals cannot be explained by simple context length or interaction progress, and they exhibit distinct formation patterns across model depth. We further show that most memory decision information is preserved in a compact recent context, while selectively restored historical evidence complements the long range dependencies that recent context misses. Based on these findings, we propose Preaction Memory with Evidence Retrieval (PaMER), which combines state guided compression with external evidence retrieval. PaMER+ further introduces step level evidence selection to recover only the historical information required by the current task. Experiments on WorkBuddyBench, across multiple context management baselines and model backbones, show that our framework substantially reduces context consumption while maintaining competitive task performance.
Modern language-model agents increasingly interact with external environments over long-horizon, multi-step trajectories, where the accumulated interaction history can quickly exceed practical context budgets. To ensure reliability, agents must maintain factual information over long horizons, remember execution errors...
Chinmay Savadikar, Zhao-Yu Zhang, Ming-Yu Zhao et al.· 0 citations
This work proposes State Conditioned Compression (StateComp), a framework that determines when historical interactions can be safely compressed according to the current agent state, and demonstrates that it reduces total agent and summarization tokens while maintaining task performance, and achieves a 12.67-fold speedu...
Large language model (LLM) agents can now undertake increasingly complex tasks, but the way they organize interaction history into memory does not ensure a coherent understanding of the current world. We introduce PoS, an inference-time framework that constructs and continually maintains explicit belief states as the a...
Yu Luo, Jia-Min Jiang, Yi-Min Zuo et al.· 1 citation
Long-horizon embodied manipulation requires policies to act on information that is no longer visible. Modified objects may leave the field of view, while completed subtasks continue to constrain future actions. Existing approaches extend temporal context through recurrent state updates, test-time training, or storage a...
Di-Qi He, Jin-Bang Guo, Yi-Jun Shen et al.· 0 citations
LLM agents have made impressive progress in coding and mathematical reasoning, motivating growing interest in learning from long-horizon interaction. Memory supports this process, but plausible lessons can be redundant or reinforce inappropriate actions. Can a memory's source decision provide a practical test for wheth...
Xin-Xiang Yin, Xing-Lang Zhang, Hao-Dong Chen et al.· 0 citations
Results characterize agent reasoning as dynamic working state rather than permanent interaction history, and suggest that historical reasoning becomes more replaceable once task relevant derived state has been reliably externalized into code, files, tool outputs, or environmental feedback.
Ming-Xuan Wang, Fei Luo, Bo Wang et al.· arXiv.org· 4 citations