AUG 18, 2026 · PREPRINT
Agentic ESOpt: Fine-Tuning Long-Horizon LLM Agents with Minimal GPU Requirements
arXiv
This is an unrefereed preprint presenting a novel algorithmic framework (Agentic ESOpt) with proof-of-concept results on a single benchmark and no peer-reviewed validation or comparison to established baselines.
Reported
Improvement over No Skill baselin…6.69%
Settings where Agentic ESOpt outp…28 of 36