EVOHARNESSBENCH: Can Your Agents Keep Pace with an Evolving Harness?

Hugging Face Daily Papers Papers

Summary

The paper introduces EVOHARNESSBENCH, a benchmark for evaluating LLM agents under evolving tool, skill, and agent harnesses, revealing gaps in retention and adaptation.

Modern LLM-based agents operate through a harness of tools, reusable skills, and specialist agents that shapes what they observe and what they can do. In practice, this harness continually evolves as new capabilities are added. We introduce EVOHARNESSBENCH, a benchmark for evaluating agents under controlled harness evolution across three axes (tools, skills, and agents). Unlike existing continual-learning benchmarks for agents, which typically place non-stationarity (i.e., what changes over time) in the task stream while keeping the harness fixed, EVOHARNESSBENCH places non-stationarity in the externally supplied harness itself. It contains 17 multi-stage harness streams constructed deterministically from verifier-based benchmarks, comprising 802 tasks, 520 tools, 42 skills, and 62 agents. We evaluate two complementary settings corresponding to the central challenges of harness evolution: deployment evaluation, which isolates retention of previously accessible competence as the harness expands, and self-evolving adaptation evaluation, which tests whether accumulated experience remains useful as new capabilities are introduced. Our results reveal three persistent gaps. First, harness expansion alone can degrade performance on previously solved tasks, producing harness-induced forgetting. Second, gains from self-evolving adaptation remain inconsistent across stages of harness evolution, capability axes, and environments. Third, retention and adaptation can pull in different directions: preserving earlier competence does not necessarily improve adaptation to newly introduced capabilities, and vice versa. These results establish harness evolution as a distinct challenge for building agents that can keep pace with an evolving harness while preserving previously effective behavior.
Original Article
View Cached Full Text

Cached at: 09/10/26, 02:15 AM

Paper page - EVOHARNESSBENCH: Can Your Agents Keep Pace with an Evolving Harness?

Source: https://huggingface.co/papers/2609.04280 Authors:

,

,

,

,

,

,

,

,

,

,

Abstract

The study introduces a benchmark to evaluate LLM agents under evolving tool, skill, and agent harnesses, revealing persistent gaps in retention, adaptation, and harness-induced forgetting.

ModernLLM-based agentsoperate through a harness of tools, reusable skills, and specialist agents that shapes what they observe and what they can do. In practice, this harness continually evolves as new capabilities are added. We introduceEVOHARNESSBENCH, a benchmark for evaluating agents under controlledharness evolutionacross three axes (tools, skills, and agents). Unlike existing continual-learning benchmarks for agents, which typically placenon-stationarity(i.e., what changes over time) in the task stream while keeping the harness fixed,EVOHARNESSBENCHplacesnon-stationarityin the externally supplied harness itself. It contains 17 multi-stage harness streams constructed deterministically from verifier-based benchmarks, comprising 802 tasks, 520 tools, 42 skills, and 62 agents. We evaluate two complementary settings corresponding to the central challenges ofharness evolution:deployment evaluation, which isolates retention of previously accessible competence as the harness expands, andself-evolving adaptationevaluation, which tests whether accumulated experience remains useful as new capabilities are introduced. Our results reveal three persistent gaps. First, harness expansion alone can degrade performance on previously solved tasks, producingharness-induced forgetting. Second, gains fromself-evolving adaptationremain inconsistent across stages ofharness evolution, capability axes, and environments. Third, retention and adaptation can pull in different directions: preserving earlier competence does not necessarily improve adaptation to newly introduced capabilities, and vice versa. These results establishharness evolutionas a distinct challenge for building agents that can keep pace with an evolving harness while preserving previously effective behavior.

View arXiv pageView PDFProject pageAdd to collection

Get this paper in your agent:

hf papers read 2609\.04280

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2609.04280 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2609.04280 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2609.04280 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Rethinking the Evaluation of Harness Evolution for Agents

arXiv cs.AI

This paper re-evaluates the methodology of automatic harness evolution for LLM agents, highlighting that its gains may stem from additional test-time search rather than improved harness design, and that evaluation on the same benchmark risks overfitting. Experiments show that harness evolution does not consistently outperform simpler test-time scaling methods.

Evo-Bench: Can Language Models Improve Agent Harness?

Hugging Face Daily Papers

Introduces Evo-Bench, the first benchmark for evaluating language models' intrinsic harness-evolving capabilities across Search, Office, and General agent domains, showing top models achieve large gains but struggle on Office workflows.

HarnessOpt-Bench: Evaluating LLMs at Harness Optimization

Hugging Face Daily Papers

HarnessOpt-Bench is a benchmark for evaluating LLMs' ability to optimize the harness—the prompts, tools, control flow, memory, and orchestration code—around a target agent, using a fixed evaluation budget. Experiments with five frontier LLMs show that optimizer models separate more than the coding harnesses they act through, with substantial room for improvement.