Hugging Face Daily PapersHaoran Ye, Yuxing Lu, Haonan DongSep 21, 20261 min readpaperadvanced

Harness-Zero: Harness Distillation via Agent-as-Harness

Summary

Harness-Zero proposes a harness‑distillation technique where a specialized harness guides a student model via an intermediate agent‑as‑harness, allowing the learned behavior to be baked into the model weights and removed at deployment. Experiments show macro‑average task success jumps from 23.3% to 44.3%, surpassing the 41.7% achieved with the harness still attached, and recovers 82.3% of harness…

Agent‑as‑harness converts optimized‑harness guidance into target‑harness action demonstrations for fine‑tuning.
Distillation yields a single model that retains specialized harness benefits without runtime overhead.
Macro‑average task success improves from 23.3% to 44.3% after distillation, exceeding the 41.7% with the harness present.
Across knowledge work, tool use, and science domains, 82.3% of 28 identified harness‑induced behavior patterns are recovered.

LLM‑based agent developers can embed complex tool‑use capabilities directly into model weights, eliminating the need for separate harness infrastructure.

7/10

llm reinforcement learning ml infra

Related reading

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 Huang9d1 minpaper
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 Li9d1 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 Ma6d2 minpaperHN12621
RetireOPD: Self-Retiring On-Policy Distillation for Agentic Reinforcement Learning arxiv.org
RetireOPD introduces a self‑retiring on‑policy distillation framework for multi‑turn RL agents. A skill‑conditioned teacher is first trained with environment rewards, then a skill‑free student learns jointly via RL and token‑level distillation. The student automatically drops the teacher once its performance gap stops shrinking and it reaches a target success‑rate fraction, after which training c…
Hugging Face Daily Papersarxiv.orgYan Yu, Zhengxi Lu, Yizhou Liu6d1 minpaper
RRSI: Regularized Recursive Self-Improvement of Agent Harnesses arxiv.org
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%.
Hugging Face Daily Papersarxiv.orgPeng Xia, Rujun Han, Zifeng Wang2d1 minpaper
Presentation: The Agent Harness: Control Planes, Invariants, and Approval Boundaries for Production AI Agents infoq.com
Production AI agents often fail silently due to issues beyond model hallucination, such as state inconsistencies. This talk outlines principles for building reliable agent harnesses, focusing on explicit state ownership, ordered mutations, and verifiable actions, using real-world case studies.
InfoQinfoq.comVinoth Govindarajan2d24 mintalk

Harness-Zero: Harness Distillation via Agent-as-Harness

Summary

Related reading

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

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

An Empirical Study of Harness Design for Coding Agentsarxiv.org

RetireOPD: Self-Retiring On-Policy Distillation for Agentic Reinforcement Learningarxiv.org

RRSI: Regularized Recursive Self-Improvement of Agent Harnessesarxiv.org

Presentation: The Agent Harness: Control Planes, Invariants, and Approval Boundaries for Production AI Agentsinfoq.com

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

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

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

RetireOPD: Self-Retiring On-Policy Distillation for Agentic Reinforcement Learning arxiv.org

RRSI: Regularized Recursive Self-Improvement of Agent Harnesses arxiv.org

Presentation: The Agent Harness: Control Planes, Invariants, and Approval Boundaries for Production AI Agents infoq.com