SkillLearnBench shows continual learning methods improve LLM agent skills over no-skill baselines but no method wins consistently across tasks and models, with external feedback helping while self-feedback causes recursive drift.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
SkillLearnBench: Benchmarking Continual Learning Methods for Agent Skill Generation on Real-World Tasks
SkillLearnBench shows continual learning methods improve LLM agent skills over no-skill baselines but no method wins consistently across tasks and models, with external feedback helping while self-feedback causes recursive drift.