将文本技能转为可插拔的权重适配器，提升大模型代理效率与安全性。
LatentSkill: From In-Context Textual Skills to In-Weight Latent Skills for LLM Agents
- 用预训练超网络将文本技能转为LoRA权重适配器，存于参数空间而非上下文
- 在ALFWorld和Search-QA上分别提升21.4和3.0分，预填充令牌减少63.9%以上
- 适配器可调参、组合，形成结构化语义空间，适合构建模块化智能体
代理系统越来越多地使用文本技能来编码可复用的任务流程，但在每一步都把技能注入提示会带来大量上下文开销，并暴露技能内容。我们提出LatentSkill，一种通过预训练超网络将文本技能转换为即插即用的LoRA适配器的框架。该方法将技能知识存储在权重空间而非上下文空间，避免了每步引入技能标记，同时保留了模块化加载、扩展和组合能力。在ALFWorld和Search-QA任务上，LatentSkill显著优于对应的上下文技能基线，平均预填充令牌减少63.9%，在已见和未见数据集上分别提升21.4和13.4分成功率；在Search-QA上精确匹配率提升3.0分，每步令牌减少71.8%。进一步分析表明，生成的技能LoRA具有结构化的语义几何，可通过缩放系数连续调节，并在技能组件对齐时通过参数空间运算实现组合。这些发现表明，权重空间技能为扩展大模型代理提供了高效、模块化且更安全的底层支持。
原文摘要 · Abstract (English)
Agent systems increasingly use textual skills to encode reusable task procedures, but injecting these skills into the prompt at every step incurs substantial context overhead and exposes skill content as plaintext. We present LatentSkill, a framework that converts textual skills into plug-and-play LoRA adapters through a pretrained hypernetwork. LatentSkill stores skill knowledge in weight space rather than context space, removing per-step skill tokens while preserving modular loading, scaling, and composition. On ALFWorld and Search-QA, LatentSkill outperforms the corresponding in-context skill baseline while using substantially fewer prefill tokens: it improves ALFWorld success by 21.4 and 13.4 points on the seen and unseen splits with 63.9% fewer prefill tokens on average, and improves Search-QA exact match by 3.0 points while using 71.8% fewer tokens per step. Further analysis shows that generated skill LoRAs form a structured semantic geometry, can be continuously modulated via the LoRA scaling coefficient, and can be composed through parameter-space arithmetic when skill components are aligned. These findings suggest that weight-space skills provide an efficient, modular, and less exposed substrate for extending LLM agents.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。