Hugging Face Daily PapersPeng Xia, Rujun Han, Zifeng WangSep 21, 20261 min readpaperadvanced

RRSI: Regularized Recursive Self-Improvement of Agent Harnesses

Summary

RRSI adds budget‑annealed edit limits, trajectory exploration incentives, a benchmark‑aware critic, and a pruning step to recursive self‑improvement of LLM agent harnesses, yielding up to +14.1 points on in‑distribution tasks and +4.7 on OOD benchmarks while cutting policy token usage by 30%.

Constraining the number of simultaneous harness edits (annealed budget) prevents over‑fitting to the training tasks.
A critic that filters proposals per benchmark and a pruner that drops tiny, expensive, or stale changes steer evolution toward reusable mechanisms.
Regularized evolution improves out‑of‑distribution performance modestly (up to 4.7 points) and reduces inference cost (≈30% fewer policy tokens).
The approach is generic: it can be plugged into any LLM‑based agent framework that supports iterative harness edits.

Recursive self‑improvement of LLM agents promises rapid capability gains, but without regularization it simply memorizes the training suite. RRSI shows a practical way to keep the evolution generalizable and efficient, a concern for anyone building production‑grade autonomous agents.

8/10

llm machine learning architecture performance ml infra

Related reading

The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement arxiv.org
The paper defines recursive self‑improvement (RSI) for AI, introduces the Headroom‑Closed Index to expose current LLM limits, and proposes a staged roadmap toward full RSI. It surveys application scenarios and outlines practical challenges.
Hugging Face Daily Papersarxiv.orgYi Duan, Ying Liu, Zirui Tang12d1 minpaperHN3
ModularRSI: Modular and Generalizable Recursive Harness Self-Improvement arxiv.org
ModularRSI proposes a contrastive, modular approach to evolve agent harnesses by independently updating five functional modules and aggregating evidence across 2,000 benchmark‑disjoint tasks. Experiments on TB2.0 and SWE‑Bench Verified show consistent gains on unseen tasks and cross‑model transfer.
Hugging Face Daily Papersarxiv.orgSiwei Wu, Jincheng Ren, Yizhi Li8d1 minpaper
Generalized Agent Iteration: One Formal Framework for Iterative Policy Improvement and Recursive Self-Improvement arxiv.org
The paper introduces Generalized Agent Iteration (GAI), a formal framework that unifies classical iterative policy improvement (GPI) and recursive self‑improvement (RSI). GAI treats an agent as a set of modifiable components and models learning as a loop of evaluation and improvement. Two binary “dials”—whether the improvement mechanism is internal to the agent and whether the evaluation standard…
Hugging Face Daily Papersarxiv.orgHongyao Tang, Yi Ma, Pengyi Li11d1 minpaper
HarnessVLN: Unifying Training-Free Embodied Navigation through an Agent Harness arxiv.org
HarnessVLN introduces a zero‑shot, training‑free embodied navigation framework that wraps a multimodal LLM in an "Agent Harness" – a tool‑based protocol that validates planner actions against spatial evidence, tracks progress with hierarchical event memory, and maintains a persistent spatiotemporal graph for recovery. The system works for instruction‑following and object‑goal tasks, achieving 60.…
Hugging Face Daily Papersarxiv.orgYang Chen, Lirong Che, Zhenyu Huang8d1 minpaper
An Empirical Study of Harness Design for Coding Agents arxiv.org
This empirical study dissects coding agent harness components like planning, action space, and context management. It finds context management is critical for tight windows, rule-based elision is efficient, and planning shifts from accuracy to cost savings for stronger models.
Hugging Face Daily Papersarxiv.orgRun-Ze Fan, Zihao Zhang, Simin Ma5d2 minpaperHN12621
SoL-Pi: Recursively Scaling Auto-Research Loops for Efficient Agent Harness arxiv.org
SoL-Pi is a new approach for scaling auto-research loops in coding agents, focusing on token efficiency for long reasoning trajectories. It achieves comparable performance to existing methods while significantly reducing token traffic by 44.7-49.0% and API costs by about one third on the 51-task EdgeBench evaluation.
Hugging Face Daily Papersarxiv.orgHaozhe Liu, Tian Ye, Sensen Gao5d1 minpaper

RRSI: Regularized Recursive Self-Improvement of Agent Harnesses

Summary

Related reading

The Last AI Built by Humans: Toward Genuine Recursive Self-Improvementarxiv.org

ModularRSI: Modular and Generalizable Recursive Harness Self-Improvementarxiv.org

Generalized Agent Iteration: One Formal Framework for Iterative Policy Improvement and Recursive Self-Improvementarxiv.org

HarnessVLN: Unifying Training-Free Embodied Navigation through an Agent Harnessarxiv.org

An Empirical Study of Harness Design for Coding Agentsarxiv.org

SoL-Pi: Recursively Scaling Auto-Research Loops for Efficient Agent Harnessarxiv.org

The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement arxiv.org

ModularRSI: Modular and Generalizable Recursive Harness Self-Improvement arxiv.org

Generalized Agent Iteration: One Formal Framework for Iterative Policy Improvement and Recursive Self-Improvement arxiv.org

HarnessVLN: Unifying Training-Free Embodied Navigation through an Agent Harness arxiv.org

An Empirical Study of Harness Design for Coding Agents arxiv.org

SoL-Pi: Recursively Scaling Auto-Research Loops for Efficient Agent Harness arxiv.org