Get Started
Research questionHow can LLM agents reuse procedural skills across diverse long-horizon tasks without generic or instance-bound memory?A single global skill document can become too generic to guide specific tasks, while storing one skill per task produces a bloated library tied to individual instances. The central difficulty is preserving reusable procedures without retaining irrelevant task-specific detail.
AI Agents
AI Memory
Reasoning
Latest papersRecent research connected to this question, newest first.SkillGLoW: Procedural-Family Skill Consolidation for Self-Improving Agents on Long-Horizon Task StreamsThe source studies LLM agents that write and reuse textual skills for mathematical reasoning, terminal automation, software repair, and embodied control. Evidence comes from four benchmarks and three models, including continual-improvement runs and transfer to unseen ALFWorld tasks; the results do not establish behavior beyond these settings.research paper · Sep 2, 2026
Related questions
How can LLM agents jointly adapt reasoning policies and hierarchical skill libraries during reinforcement learning?How can computer-use agents retain, refine, and reliably reuse procedural skills across repeated GUI tasks?How can LLM agents retrieve relevant skills from large, noisy libraries under context and latency constraints?How can LLM agents reuse execution traces without losing temporal and outcome-dependent behavior?
Home
Topics
Search
Library