The paper introduces Neural Procedural Memory (NPM), a training-free framework that enables Large Language Model agents to utilize implicit activation steering for procedural memory instead of relying on explicit textual instructions. By distilling skills from historical experiences into steering vectors, NPM directly activates task-relevant neural mechanisms to guide execution.
- NPM represents agent memory through implicit activation steering rather than Retrieval-Augmented Generation (RAG) guidelines.
- Procedural skills are distilled from historical contrastive experiences into steering vectors within the activation space.
- Evaluations across four agent benchmarks show performance comparable to baselines using explicit textual instructions.
- Combining implicit steering with explicit workflows provides complementary advantages for more robust task execution.
- Representational analyses indicate that steering vectors encode consistent task logic and form organized structures in the activation space.
The authors consider this significant because it addresses the text-action disconnect found in symbolic instruction methods, offering a promising approach for managing agent memory through direct neural mechanism activation.