Organizations: Beijing Institute of Technology · Tsinghua University · Chengdu Institute of Computer Applications, Chinese Academy of Sciences · University of Chinese Academy of Sciences · Beijing University of Technology · Yangtze Delta Region Institute of Tsinghua University, Zhejiang · Jiaxing Key Laboratory of Artificial Intelligence and Cyber Resilience
Recursive self-improvement (RSI) of agent harnesses has emerged as an effective approach to enhancing the capabilities of large language model (LLM) agents. By searching the harness space for a specific LLM on a given dataset, RSI can discover a harness tailored to the capabilities and behavioral characteristics of that model. However, such optimization often creates a strong coupling between the learned harness and its target LLM, limiting generalization across models and requiring substantial computational resources to adapt to new LLMs. Moreover, existing RSI methods typically optimize a harness using feedback from a single model and evaluate its generalization to other models only after training. This paradigm treats model differences primarily as a criterion for evaluating generalization rather than as valuable evidence for optimization. In this paper, we introduce a new perspective that treats models as data and places the harness at the center of optimization. Building on this perspective, we propose MaD-RSI, a framework that directly optimizes a shared harness across multiple heterogeneous LLMs by incorporating their diverse failure trajectories into the training process. Rather than assessing model differences only after optimization, MaD-RSI leverages these differences as training evidence to guide the self-improvement of a shared harness. Furthermore, we introduce a new evaluation framework with two complementary metrics designed to assess the effectiveness of harness optimization across models. Experiments across multiple datasets and LLMs demonstrate that MaD-RSI improves generalization over existing approaches while reducing training costs.
Figures & tables
Figure 1: Comparison between MaD-RSI and existing RSI approaches that evolve agent harnesses using feedback from a single backbone model.
Figure 2: Comparison of MaD-RSI against four independently optimized baselines in performance and resource costs. Safety performance is measured by UA.
Scheme
LawBench
MBPP+
MedXpertQA
Pass@4
Pass^4
Pass@4
Pass^4
Pass@4
Pass^4
Original
62.50
42.50
93.00
76.50
60.00
13.00
RRSI / DeepSeek-V4
61.50
41.50
93.00
77.50
60.00
13.00
RRSI / DeepSeek-V3.2
61.00
42.50
92.50
78.50
54.50
11.50
RRSI / Qwen3.8-Flash
61.00
43.50
91.50
77.50
55.00
12.00
RRSI / Qwen3.8-27B
60.50
41.50
92.00
78.50
52.50
11.50
Table 1: Four-model evaluation across general and safety tasks.
Dataset
Metric
Pass@6
Pass^6
Baseline
MaD-RSI
Baseline
MaD-RSI
LawBench
Accuracy
61.00
73.00
36.00
38.50
MBPP+
94.00
94.00
74.50
77.50
MedXpertQA
67.00
67.00
8.50
13.00
SafetyBench
AS
100.00
100.00
94.64
97.50
UA
96.36
95.45
66.36
76.36
Table 2: Expanded six-model evaluation. The population contains four development executors plus Doubao-2.0 and GLM-5.3-Flash.
An LLM agent's capability is largely magnified by its harness, namely the prompts, control flow, tooling, memory, and context management surrounding the frozen backbone model. Recent methods increasingly automate this process by iteratively proposing and selecting component-wise edits of an agent harness, practically establishing a form of recursive self-improvement (RSI) at the agent-system level. However, such recursive evolution may overfit by memorizing the training tasks, showing large in-distribution gains that shrink or even vanish on out-of-distribution benchmarks. We introduce Regularized Recursive Self-Improvement of Agent Harnesses (RRSI), which incorporates the principles of regularizations into harness self-improvement by constraining the evolution candidate proposal and selection. The proposer operates with a temporally annealed budget, limiting how many edits a candidate can bundle, and it encourages unexplored trajectories based on evolution history. The selector is equipped with a critic and a pruner: the critic screens benchmark-specific proposals, while the pruner, removes changes that are too small, too expensive, or no longer useful. Together these constraints favor reusable agent mechanisms over benchmark-specific ones or even noises. Across eight benchmarks spanning coding, agentic workspace and engineering design tasks, RRSI gains up to 6.0 points on the split it evolves against and up to 4.3 points on the five out-of-distribution benchmarks, while producing a harness that runs on 30% fewer policy tokens than the unregularized evolution. Code is available at https://github.com/google-research/rrsi and project page is https://regularized-rsi.com/.
Peng Xia, Rujun Han, Zifeng Wang +11
Google Cloud AI Research · UNC-Chapel Hill · Stanford University +1
The performance of LLM-based agents is jointly shaped by their base models and the harnesses that mediate their interaction with the environment. Because different models exhibit distinct behaviors, effective harness design is inherently model-specific. Yet agent harnesses are still largely engineered by human experts, a paradigm that scales poorly as modern LLMs become increasingly diverse and rapidly evolving. In this paper, we introduce Self-Harness, a new paradigm in which an LLM-based agent improves its own operating harness, without relying on human engineers or stronger external agents. We operationalize Self-Harness as an iterative loop with three stages: Weakness Mining, which identifies model-specific failure patterns from execution traces; Harness Proposal, which generates diverse yet minimal harness modifications tied to these failures; and Proposal Validation, which accepts candidate edits only after regression testing. We instantiate Self-Harness on Terminal-Bench-2.0 using a minimal initial harness and three base models from diverse families: MiniMax M2.5, Qwen3.5-35B-A3B, and GLM-5. Across all three models, Self-Harness consistently improves performance, with held-out pass rates increasing from 40.5% to 61.9%, 23.8% to 38.1%, and 42.9% to 57.1%, respectively. Qualitative analyses further show that Self-Harness does not simply add generic instructions, but effectively turns model-specific weaknesses into concrete, executable harness changes. These results suggest a path toward LLM-based agents that are not merely shaped by their harnesses, but can also participate in reshaping them.
Autonomous agents increasingly depend on harnesses that govern interaction, memory, tools, and execution around large language models (LLMs). While recent work has begun to evolve these harnesses from execution experience, the supervision used to guide evolution often remains relatively static, making it difficult to track changing behavioral weaknesses across rounds and potentially limiting generalization to unseen tasks. To address this issue, we propose RuHarness, a rubric--harness co-evolution framework for continual agent self-improvement. RuHarness continually discovers and reassesses behavioral criteria from execution trajectories, prioritizes the most relevant unresolved weaknesses, and uses them to guide targeted harness updates while keeping the task objective fixed. Extensive experiments across diverse agent tasks and backbone models show that RuHarness outperforms strong harness-evolution baselines, particularly on unseen held-out tasks. Further analyses show that adaptive rubric prioritization tracks shifting weaknesses and supports more sustained improvement and stable generalization across evolution rounds.
Hao Chen, Yukun Yan, Ziyu Han +7
Research Center for Social Computing and Interactive Robotics, Harbin Institute of Technology · Department of Computer Science and Technology, Institute for AI, Tsinghua University