Back to feed
arXiv cs.AI·

Library Drift: Diagnosing and Fixing a Silent Failure Mode in Self-Evolving LLM Skill Libraries

Signal
78
Hype
15
In three linesSelf-evolving skill libraries suffer silent degradation termed 'library drift': unbounded accumulation without lifecycle management. Study isolates mechanism via ablations, provides trace-level diagnostics, and validates fix (outcome-driven retirement + bounded active-cap + meta-skill prior) lifting pass@1 from 0.258 baseline to 0.584 on MBPP+ hard-100.
Read source
Your take?
AI AgentsCode generationBenchmarksReinforcement learning

Summary generated by Claude — human-verified