HyperAIHyperAI

Command Palette

Search for a command to run...

Agentic ESOpt: 最小限のGPUメモリ要件による長時間水平線LLMエージェントのファインチューニング

Zhi Zheng Rongsheng Chen Yunpeng Ba Zhenkun Wang Yee Whye Teh Wee Sun Lee

概要

強化学習(RL)は単一ターンのLLMファインチューニングにおいて有望視されてきた。しかし、長時間水平線のエージェント的推論は分岐する相互作用とスパースな報酬を増大させ、RLのいくつかの限界を露呈させる。すなわち、重い誤差逆伝播により大規模LLMのファインチューニングが非現実的になること、そして長時間水平線の軌跡によりクレジット割り当てが著しく困難になることである。本論文は、長時間水平線LLMエージェントのファインチューニングにおいて、進化戦略(ES)がより優れた選択肢となり得ると論じる。エージェント的RLと比較して、ESは3つの重要な利点を提供する。1)モデル拡張性:ESは推論レベルの最小限のGPUメモリのみで全パラメータ最適化を可能にし、大規模LLMのファインチューニングを実現する。2)柔軟性:軽量なブラックボックスフィードバックインターフェースにより、ESファインチューニングはプロンプト空間の進化(例:スキル最適化やテスト時計算)と容易に組み合わせられる。3)長時間水平線拡張性:ESは水平線全体に報酬を分解することなく軌跡レベルのパラメータ帰属を行い、水平線長の増大に伴いエージェント的RLよりも優れた拡張性を示す。この知見に基づき、我々は柔軟なパラメータ・コンテキスト共進化に特化した全パラメータエージェント的ファインチューニングフレームワークであるAgentic ESOptを提案する。各ステップにおいて、Agentic ESOptは現在のLLMパラメータ周辺に摂動をサンプリングし、得られたエージェントを報酬で評価し、オンライン報酬重み付き更新を適用する。探索と適応のトレードオフを改善するため、Agentic ESOptはさらに摂動スケールσのコサイン減衰スケジュールを導入する。我々はAgentic ESOptを訓練時ファインチューニングとエージェント的テスト時計算の両方の設定で評価する。長時間水平線の数独において、Agentic ESOptはQwen3.5-4BでRL手法を12.50%上回る。WebArena-Liteでは、Qwen3.5-27Bの全パラメータ最適化がNo Skillベースラインを6.69%改善し、Agentic ESOptとTrace2Skillの組み合わせはTrace2Skillベースラインをさらに2.42%改善する。テスト時自動ヒューリスティック設計において、Agentic ESOptはオンラインのプロンプト・パラメータ共進化を実行し、36設定中28設定で対応するベースラインを改善する。

One-sentence Summary

Researchers from National University of Singapore, Southern University of Science and Technology, and Oxford propose Agentic ESOpt, a full-parameter evolution-strategy fine-tuning framework for long-horizon LLM agents that uses perturbation sampling, online reward-weighted updates, and a cosine decay schedule for the perturbation scale σ while requiring only inference-level GPU memory; on long-horizon Sudoku it outperforms RL by 12.50% with Qwen3.5-4B, and on WebArena-Lite full-parameter optimization of Qwen3.5-27B improves the No Skill baseline by 6.69% while combining Agentic ESOpt with Trace2Skill improves the Trace2Skill baseline by 2.42%.

Key Contributions

  • Agentic ESOpt is introduced as a full-parameter evolution-strategy fine-tuning framework for long-horizon LLM agents that samples parameter perturbations, evaluates the resulting agents with trajectory-level rewards, and applies online reward-weighted updates without backpropagation or explicit per-action credit assignment, requiring only inference-level GPU memory.
  • The method includes a cosine decay schedule for the perturbation scale to manage the exploration-adaptation trade-off and supports parameter-context co-evolution with prompt-space optimization such as skill optimization and test-time compute.
  • Experiments show that Agentic ESOpt outperforms agentic RL baselines on long-horizon Sudoku by 12.50% with Qwen3.5-4B; on WebArena-Lite, full-parameter optimization of Qwen3.5-27B improves the No Skill baseline by 6.69%, and combining Agentic ESOpt with Trace2Skill improves the Trace2Skill baseline by 2.42%; in test-time automatic heuristic design, Agentic ESOpt improves its matched baseline in 28 of 36 settings. Its inference-level memory footprint enables full-parameter adaptation of a Qwen3.5-27B WebArena agent on four H100 GPUs.

Introduction

Advanced Large Language Models such as Qwen3, DeepSeek-R1, and Gemini 2.5 can act as general-purpose agents that use tools, process long contexts, and coordinate multi-step software workflows, but they still struggle with uncommon tool APIs and specialized scientific or algorithmic tasks, so efficient task-specific fine-tuning remains important. Prior agentic reinforcement learning methods require heavyweight activations and optimizer states, backpropagation through long branching trajectories, and turn-level or critic-based credit assignment under sparse rewards, which becomes increasingly impractical as horizons grow. Previous evolution-strategy work on single-turn reasoning was mostly memory efficient but often slightly weaker than RL. The authors argue that for long-horizon agentic reasoning, evolution strategies become structurally preferable rather than merely cheaper. They introduce Agentic ESOpt, a full-parameter evolution-strategy framework that samples parameter perturbations, evaluates agents with environment rewards, and applies reward-weighted updates without backpropagation. It supports both train-time fine-tuning and test-time compute, uses a cosine perturbation schedule to balance exploration and adaptation, and enables full-parameter optimization with only inference-level GPU memory.

Method

The authors propose Agentic ESOpt, a method that performs full-parameter Evolutionary Strategy (ES) optimization for Large Language Models (LLMs). The framework operates by sampling parameter perturbations around the current LLM, evaluating the perturbed agents using scalar environment rewards, and applying a reward-weighted parameter update. This forward-only process requires storing only the noise seed and utilizing in-place addition and subtraction, ensuring that GPU memory requirements remain minimal and equivalent to standard inference. Additionally, the black-box trajectory feedback generated can be reused by skill-space optimizers or test-time compute procedures.

As illustrated in the framework diagram:

Formally, let τ=(o0,a0,,oH,aH)\pmb{\tau} = (o_0, a_0, \dots, o_H, a_H)τ=(o0,a0,,oH,aH) denote an interaction trajectory induced by policy πθ\pi_\thetaπθ, and let R(τ)R(\tau)R(τ) denote its scalar trajectory return. For a fixed external agent state ccc, the objective is defined as:

J(θ;c)=Eτπθ(c)[R(τ)]J(\theta; c) = \mathbb{E}_{\boldsymbol{\tau} \sim \pi_\theta(\cdot|c)} [R(\boldsymbol{\tau})]J(θ;c)=Eτπθ(c)[R(τ)]

Agentic ESOpt optimizes this objective by searching the parameter space around θ\thetaθ. Let ϵRd\epsilon \in \mathbb{R}^dϵRd be a full-parameter perturbation on the ddd-dimensional parameters. The Gaussian-smoothed objective is formulated as:

Jσ(θ;c)=EϵN(0,I)[J(θ+σϵ;c)]J_\sigma(\theta; c) = \mathbb{E}_{\epsilon \sim \mathcal{N}(0, I)} [J(\theta + \sigma \epsilon; c)]Jσ(θ;c)=EϵN(0,I)[J(θ+σϵ;c)]

The ES pseudo-gradient is then derived as:

θJσ(θ;c)=1σEϵ[J(θ+σϵ;c)ϵ]\nabla_\theta J_\sigma(\theta; c) = \frac{1}{\sigma} \mathbb{E}_\epsilon [J(\theta + \sigma \epsilon; c) \epsilon]θJσ(θ;c)=σ1Eϵ[J(θ+σϵ;c)ϵ]

These ES gradients are estimated from scalar scores without differentiating through the agent-environment interaction. In implementation, to estimate the gradient, the authors sample GGG perturbations ϵ1,,ϵG\epsilon_1, \dots, \epsilon_Gϵ1,,ϵG, evaluate the corresponding perturbed agents, and obtain rewards Ri=R(τi)R_i = R(\tau_i)Ri=R(τi). To reduce variance, the rewards are normalized within the population using a z-score:

R^i=RiμRsR+ε,whereμR=1Gj=1GRj,sR2=1Gj=1G(RjμR)2\hat{R}_i = \frac{R_i - \mu_R}{s_R + \varepsilon}, \quad \text{where} \quad \mu_R = \frac{1}{G} \sum_{j=1}^G R_j, \quad s_R^2 = \frac{1}{G} \sum_{j=1}^G (R_j - \mu_R)^2R^i=sR+εRiμR,whereμR=G1j=1GRj,sR2=G1j=1G(RjμR)2

In practice, the explicit 1/σ1/\sigma1/σ factor is omitted, with α\alphaα serving as the effective update scale. The implemented update rule is:

θt+1=θt+αGi=1GR^iϵi\theta_{t+1} = \theta_t + \frac{\alpha}{G} \sum_{i=1}^G \hat{R}_i \boldsymbol{\epsilon}_iθt+1=θt+Gαi=1GR^iϵi

The method also facilitates prompt-space composition and prompt-parameter co-evolution. Unlike test-time compute and prompt-space optimization methods that keep LLM parameters fixed, the lightweight black-box updates of Agentic ESOpt allow parameter adaptation to occur on the fly alongside prompt-space search. Let Dt\mathcal{D}_tDt denote the trajectories and scores collected at iteration ttt, UES\mathcal{U}_{ES}UES the Agentic ESOpt parameter update, and Uc\mathcal{U}_cUc an external update rule for prompt ctc_tct. A general alternating outer loop updates the two spaces as:

θt+1=UES(θt;ct,Dt),ct+1=Uc(ct;Dt)\theta_{t+1} = \mathcal{U}_{ES}(\theta_t; c_t, \mathcal{D}_t), \quad c_{t+1} = \mathcal{U}_c(c_t; \mathcal{D}_t)θt+1=UES(θt;ct,Dt),ct+1=Uc(ct;Dt)

To manage the trade-off between exploration and objective bias, the authors introduce a cosine decay for the perturbation radius σ\sigmaσ. The Gaussian-smoothed objective introduces a smoothing bias characterized by the second-order term Tr(θ2J)\text{Tr}(\nabla_\theta^2 J)Tr(θ2J), which acts as a regularization term penalizing sharp local optima. A larger σ\sigmaσ increases this regularization but also increases bias from the original objective. The radius is gradually decreased over TTT update steps as follows:

σt=σT+(σ0σT)1+cos(πt/T)2,t=0,,T\sigma_t = \sigma_T + (\sigma_0 - \sigma_T) \frac{1 + \cos(\pi t / T)}{2}, \qquad t = 0, \dots, Tσt=σT+(σ0σT)21+cos(πt/T),t=0,,T

For train-time Agentic ESOpt, a nonzero σT\sigma_TσT is retained to balance exploitation with exploration and regularization. In contrast, for test-time compute, which focuses on the unbiased outcome of the current task rather than generalization, σT\sigma_TσT is decayed to zero to minimize objective bias toward the end of optimization.

Experiment

The experiments evaluate Agentic ESOpt across controlled long-horizon tasks, train-time fine-tuning, test-time compute, and population scaling. In multi-turn Sudoku, parameter-space search shows a horizon-dependent advantage over action-space policy gradients because it avoids per-step credit assignment, while also requiring only inference-level GPU memory. For ReAct-style Math and DocVQA fine-tuning, Agentic ESOpt consistently outperforms matched Agentic GRPO baselines and composes effectively with Trace2Skill, and WebArena results confirm that full-parameter adaptation of a 27B agent is feasible. On automatic heuristic design, Agentic ESOpt improves most Sample and EoH comparisons under matched budgets, and population-scaling experiments suggest stronger backbones are less sensitive to small population sizes.

Success rates decline for all methods as the minimum required planning horizon increases. Agentic ESOpt with sigma decay maintains the highest success at the longest horizon and keeps GPU memory at the lightweight 4B backbone level, while Agentic PPO collapses to zero success and requires substantially more memory. Removing sigma decay from the ES run reduces success relative to Agentic ESOpt across tested horizons. Every evaluated method loses success rate as the minimum successful horizon grows. Agentic ESOpt is the strongest at the longest horizon and uses far less GPU memory than Agentic PPO and Agentic GRPO. Agentic PPO performs well at the shortest horizon but fails at the longest horizon. Vanilla ES without sigma decay trails Agentic ESOpt at all tested horizons.

Across the agentic Sudoku horizons tested, Qwen3.5-4B with Agentic ESOpt requires less training compute and less wall-clock time than with Agentic GRPO on the same four-GPU hardware. The compute advantage is small at the shortest horizon and widens as the horizon lengthens, while wall-clock time remains substantially lower throughout. Agentic ESOpt reduces training FLOPs relative to Agentic GRPO at every tested horizon, with the savings increasing as the horizon grows. Wall-clock time for Agentic ESOpt is roughly half that of Agentic GRPO at the longer horizons and remains lower at the shortest horizon, despite both methods fully using the same four H100 GPUs.

Agentic ESOpt consistently improves the Qwen3.5-4B base model and matched Agentic GRPO baselines on math reasoning and DocVQA. The gains are strongest on DAPO, AIME 2026, and DocVQA Mean@4, and the method composes with Trace2Skill to reach the strongest Qwen3.5-4B Mean@4 results. Pass@4 and Max@4 metrics also favor Agentic ESOpt over Agentic GRPO, while DAPO Pass@4 is unchanged from the No Skill baseline. Agentic ESOpt plus No Skill produces double-digit Mean@4 gains on DAPO and AIME 2026 and a large DocVQA accuracy improvement over the Qwen3.5-4B No Skill baseline. Agentic ESOpt plus No Skill outperforms matched Agentic GRPO plus No Skill on DAPO, AIME 2026, and DocVQA mean and best-of-four metrics, with especially large AIME 2026 Pass@4 and DocVQA Max@4 gains.

On WebArena-Lite, Agentic ESOpt improves the Qwen3.5-27B No Skill baseline on the dataset average and across most major site categories, with particularly large gains in OSS, GitLab, CMS, and Map. The Reddit category is a slight exception, where the adapted model does not improve over its baseline. Combining Agentic ESOpt with Trace2Skill yields the highest reported Qwen3.5-27B average, indicating compatibility with skill-space optimization. Agentic ESOpt improves the Qwen3.5-27B No Skill baseline on the dataset average and in four of the five major site categories, while Reddit shows a small decline. The largest category gains over the No Skill baseline appear in OSS, GitLab, CMS, and Map, in that order. The strongest Qwen3.5-27B dataset average is achieved by combining Agentic ESOpt with Trace2Skill, surpassing the No Skill and Trace2Skill baselines.

Agentic ESOpt improves existing heuristic design search methods across most constructive problem settings. When paired with EoH, it improves all six constructive test sets at both evaluation budgets; when paired with Sample, it improves nine of twelve matched comparisons. Across constructive and ACO-style settings, the approach improves 28 of 36 matched comparisons. Agentic ESOpt + EoH improves all six constructive heuristic design test sets at both evaluation budgets. Agentic ESOpt + Sample improves nine of twelve matched comparisons, with one tie and two regressions. Across both constructive baselines, Agentic ESOpt improves 21 of 24 matched comparisons. In the T=1000 results, Agentic ESOpt + Sample shows the largest relative gains on TSP instances and smaller mixed changes on KP and ASP instances.

Across agentic Sudoku, math reasoning, web navigation, and heuristic design tasks, Agentic ESOpt is evaluated against PPO, GRPO, and baseline skill or search methods. It maintains higher success at long planning horizons while using substantially less GPU memory, training compute, and wall-clock time than PPO and GRPO, and sigma decay contributes to that advantage. Agentic ESOpt also improves Qwen3.5-4B results on math reasoning and DocVQA, improves Qwen3.5-27B WebArena-Lite performance especially when combined with Trace2Skill, and enhances existing heuristic design search methods in most matched comparisons.


AIでAIを構築

アイデアからローンチまで — 無料のAIコーディング支援、すぐに使える環境、最高のGPU価格でAI開発を加速。

AI コーディング補助
すぐに使える GPU
最適な料金体系

HyperAI Newsletters

最新情報を購読する
北京時間 毎週月曜日の午前9時 に、その週の最新情報をメールでお届けします
メール配信サービスは MailChimp によって提供されています