Loading
Life sciences · Preprint
arXiv · August 18, 2026
Early or partial results. Treat as a signal, not a conclusion.
This preprint proposes Agentic ESOpt, an evolution-strategies-based framework for full-parameter fine-tuning of long-horizon LLM agents, claiming advantages over reinforcement learning in scalability and credit assignment. Empirical results show a 6.69% improvement over a no-skill baseline on WebArena-Lite and outperformance in 28 of 36 test-time co-evolution settings, but lack peer review, independent validation, and rigorous baseline comparisons necessary to assess true impact.
Preprint. Large language models (Qwen-3.5-27B evaluated); agentic tasks requiring long-horizon reasoning on WebArena-Lite benchmark.. Intervention: Agentic ESOpt: evolution-strategies-based full-parameter fine-tuning with parameter perturbation sampling, reward-weighted updates, and cosine decay schedule for perturbation scale; includes test-time prompt–parameter co-evolution.. Compared with: No Skill baseline on WebArena-Lite; matched baselines in test-time heuristic design (36 settings).
Full-parameter optimization of Qwen-3.5-27B on WebArena-Lite improves No Skill baseline by 6.69% In test-time automatic heuristic design, Agentic ESOpt outperforms matched baseline in 28 of 36 settings
Safety was not reported in the material analysed. Check the source before drawing any conclusion about harm.
The source did not state who this applies to in practice.
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.
As stated by the source record.
Quoted from the source exactly as published.
Graded across the dimensions that decide whether you should act, each from what the source actually supports. There is no single score, and where a dimension was not assessed it says so.
Reinforcement Learning (RL) has been promising in single-turn LLM fine-tuning. However, long-horizon agentic reasoning introduces increasingly branching interactions and sparse rewards, exposing several limitations of RL: its heavyweight backpropagation-based training stack makes it impractical to fine-tune larger LLMs, and longer-horizon trajectories make credit assignment in RL substantially harder. This paper argues that evolution strategies (ES) can be a better choice for fine-tuning long-horizon LLM agents. Compared with agentic RL, ES offers three key advantages: 1) Model Scalability: ES enables full-parameter optimization with only minimal, inference-level GPU memory, making it possible to fine-tune large LLMs. 2) Flexibility: its lightweight, black-box feedback interface makes ES fine-tuning easy to compose with prompt-space evolution (e.g., skill optimization & test-time compute); and 3) Long-Horizon Scalability: ES performs trajectory-level parameter attribution without decomposing rewards across horizons, yielding better scalability than Agentic RL as the horizon length grows. Based on this insight, we propose Agentic ESOpt, a full-parameter agentic fine-tuning framework tailored to flexible parameter--context co-evolution. At each step, Agentic ESOpt samples perturbations around the current LLM parameters, evaluates the resulting agents with rewards, and applies an online reward-weighted update. To improve the exploration--adaptation trade-off, Agentic ESOpt further introduces a cosine decay schedule of the perturbation scale $σ$. On WebArena-Lite, full-parameter optimization of Qwen-3.5-27B improves the No Skill baseline by 6.69%. In test-time automatic heuristic design, Agentic ESOpt performs online prompt--parameter co-evolution, improving its matched baseline in 28 of 36 settings.
Taken from the source record, never inferred. Follow any of these and new work involving them reaches your briefing.