analiza aiqbench
HarnessX addresses a critical bottleneck in AI agent development by automating the evolution of runtime harnesses—such as prompts and control flows—using execution traces rather than relying solely on static hand-crafting. The system demonstrates that significant performance gains can be achieved through adaptive scaffolding, offering a viable alternative to pure model scaling for improving task completion rates across diverse benchmarks. This work highlights an actionable path forward where agent progress is driven by systematic feedback loops between runtime behavior and interface composition.
HarnessX: A Composable, Adaptive, and Evolvable Agent Harness Foundry
arXiv:2606.14249v1 Announce Type: new Abstract: AI agent performance depends critically on the runtime harness, comprising the prompts, tools, memory, and control flow that mediate how a model observes, reasons, and acts. Yet today's harnesses remain largely hand-crafted and static: each new model or task still demands bespoke scaffolding, and the rich traces produced during execution are rarely distilled back into systematic improvement. We introduce HarnessX, a foundry for composable, adaptive, and evolvable agent harnesses. HarnessX assembles typed harness primitives via a substitution algebra, adapts them through AEGIS, a trace-driven multi-agent evolution engine grounded in an operational mirror between symbolic adaptation and reinforcement learning, and closes the harness-model loop by turning trajectories into both harness updates and model training signal. Across five benchmarks (ALFWorld, GAIA, WebShop, tau^3-Bench, and SWE-bench Verified), HarnessX yields an average gain of +14.5% (up to +44.0%), with gains largest where baselines are lowest. These results suggest that agent progress need not come from model scaling alone: composing and evolving runtime interfaces from execution feedback is an actionable and complementary lever. The complete codebase will be open-sourced in a future release.
Abstract:AI agent performance depends critically on the runtime harness, comprising the prompts, tools, memory, and control flow that mediate how a model observes, reasons, and acts. Yet today's harnesses remain largely hand-crafted and static: each new model or task still demands bespoke scaffolding, and the rich traces produced during execution are rarely distilled back into systematic improvement. We introduce HarnessX, a foundry for composable, adaptive, and evolvable agent harnesses. HarnessX assembles typed harness primitives via a substitution algebra, adapts them through AEGIS, a trace-driven multi-agent evolution engine grounded in an operational mirror between symbolic adaptation and reinforcement learning, and closes the harness-model loop by turning trajectories into both harness updates and model training signal. Across five benchmarks (ALFWorld, GAIA, WebShop, tau^3-Bench, and SWE-bench Verified), HarnessX yields an average gain of +14.5% (up to +44.0%), with gains largest where baselines are lowest. These results suggest that agent progress need not come from model scaling alone: composing and evolving runtime interfaces from execution feedback is an actionable and complementary lever. The complete codebase will be open-sourced in a future release.
Artificial Intelligence (cs.AI)
arXiv:2606.14249 [cs.AI]
(or arXiv:2606.14249v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2606.14249 arXiv-issued DOI via DataCite (pending registration)
Text integral sub licența deschisă a publicației · 196 cuvinte
Subiecte conexe
HarnessX: A Composable, Adaptive, and Evolvable Agent Harness Foundry
arXiv · 24 iul. 2026
openJiuwen: Beyond Static Harnesses for Long-Horizon Coding Agents
arXiv · 1 sept. 2026
EvolveNet: Collaborative Harness Evolution for Agent Self-Improvement
arXiv · 6 aug. 2026
LiteEvo: Automated, Cost-Efficient Harness Evolution for Generalization to Unseen Tasks
arXiv · 29 sept. 2026