Coca-Cola：面向LLM智能体的潜空间上下文压缩方法
Tools Need Not Speak in Text: Latent Context Compression for Agents
AI 导读
论文提出 Coca-Cola，一种面向 LLM 智能体的连续潜空间上下文压缩模型，将工具观测压缩为潜 token 嵌入智能体输入空间，并引入 Action-Preserving Training 对齐全量与压缩观测下的下一步动作分布。
正文 · 原文
Abstract:Large language model (LLM) agents typically append raw tool outputs to their context. Over multi-step interactions, long and noisy observations incur substantial context overhead and dilute the agent's attention to task-relevant information. Existing token pruning and summarization methods remain constrained by discrete text representations, limiting the achievable compression ratios.
In this work, we challenge the prevailing assumption that agents must consume tool observations as explicit text tokens. To this end, we propose Coca-Cola, a continuous latent context compression model for LLM agents. Coca-Cola compresses tool observations into latent tokens and embeds them into the input embedding space of the agent. We further introduce Action-Preserving Training, which aligns the agent's next-action distributions under full and compressed observations. This trains the compression model to learn action-sufficient latent tokens that retain decision-critical information. Experiments with web-search agents on four QA benchmarks and two model families show that Coca-Cola reduces the agent's token consumption by 70.01%-89.24% and end-to-end runtime by 24.45%-55.97%, while consistently improving task performance. These results suggest that Coca-Cola offers a promising approach for context compression in agents. Our code is available at this https URL.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.06500 [cs.AI] |
| (or arXiv:2611.06500v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2611.06500 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Yuetong Liu [view email]
[v1]
Wed, 4 Nov 2026 13:32:36 UTC (1,232 KB)
来源：arXiv cs.AI · arxiv.org