RobustSGPO: Search-Space Control for Agent Harness Evolution

• URL: http://arxiv.org/abs/2609.09646v1
• Date: Wed, 09 Sep 2026 03:00:03 GMT
• Title: RobustSGPO: Search-Space Control for Agent Harness Evolution
• Abstract summary: We introduce RobustSGPO, which specifies the requested edit, constructs and checks the patch, and continues search from either the incumbent or retained snapshots.<n>We evaluate permission scheduling, cumulative controls, and task-family transfer in the AgentX brainstorming workflow using 120 tasks, 95 runs, and 7,350 candidate attempts.
• Score: 56.722805974785125
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Semantic-gradient-based prompt optimization (SGPO) improves agent harnesses using execution feedback, but its local update rule leaves the choice of edit scope and operation unresolved. We introduce RobustSGPO, which specifies the requested edit, constructs and checks the patch, and continues search from either the incumbent or retained snapshots. We evaluate permission scheduling, cumulative controls, and task-family transfer in the AgentX brainstorming workflow using 120 tasks, 95 runs, and 7,350 candidate attempts. Periodic $1\to2\to3$ scheduling exceeds fixed maximum permission by 0.28 test-score points. RobustSGPO increases completion on 30 held-out tasks from 60.0% to 80.0% and improves test quality from 3.77 to 4.14 under a 20-million-token budget. Category retention reduces source-task degradation after a shift, whereas random retention reaches a higher destination endpoint. Search-space control benefits quality through executable edits and alternative starting points, with measurable retention overhead.

Related papers

• Task-Conditioned Least-Privilege Learning for Executable Terminal and MCP Agents [45.88028371034407]
  Tool-using large language-model agents can complete a task while exercising authority that the user did not grant or the task does not need.<n>We propose a framework where each action is audited before execution and again from observed effects along six dimensions of risk.<n>We find that after training using this framework on Qwen3.5-4B over 1,500 tasks, the selected seed reaches 98.48% safe success across 2,896 evaluation episodes.
  arXiv  Detail & Related papers  (2026-08-18T22:07:23Z)
• DG-FedReuse: Proxy-Gradient-Gated Cached-Update Reuse with Matched Sparse Uplink Accounting [0.0]
  DG-FedReuse is a simulator-level mechanism that allows selected clients to contribute cached updates.<n>Experiments cover six image-classification datasets, 50 virtual clients, Dirichlet label clients, and three seeds.<n> DG-FedReuse yields 83.36-85.42% modeled update-data-field uplink saving, compared with 76.88% for matched Top-K FedAvg.
  arXiv  Detail & Related papers  (2026-08-05T19:30:04Z)
• Fail-Fast, Restart-Smart: Early Failure Prediction and Restart for SWE Agentic Tasks [11.325524635118846]
  FailFast-RestartSmart is a two-stage controller for a single active trajectory.<n>It predicts failure from observable prefixes without policy logits or hidden states.<n>Upon an alarm, RestartSmart launches a fresh same-policy rollout without prior prompt history.
  arXiv  Detail & Related papers  (2026-08-04T06:55:46Z)
• A Self-Triggered Agentic Push Recommendation System [77.13502692161426]
  Push notification is a critical recommendation scenario on large-scale platforms.<n>We propose a proactive, Self-Triggered End-to-end Agentic Push Recommendation System.<n>STEPS is already fully deployed at Douyin with over 1 billion users.
  arXiv  Detail & Related papers  (2026-08-03T09:20:53Z)
• Do Agent Optimizers Compound? A Continual-Learning Evaluation on Terminal-Bench 2.0 [53.71882250666667]
  Most reported gains from agent-optimization methods are one-shot.<n>This does not test the setting that matters for deployed agents.<n>We compare three approaches to agent-harness optimization.
  arXiv  Detail & Related papers  (2026-07-15T16:36:04Z)
• Mastermind: Strategy-grounded Learning for Repository-Scale Vulnerability Reproduction [62.42945715963878]
  Repository-level vulnerability reproduction is a demanding software engineering (SE) task.<n>Recent LLM agents can often execute these steps when the approach is correct, yet they still fail by choosing the wrong strategy.<n>This paper argues that strategy, rather than the full action trajectory, is the right learning unit for such SE agents.
  arXiv  Detail & Related papers  (2026-07-02T06:27:27Z)
• PACE: Anytime-Valid Acceptance Tests for Self-Evolving Agents [0.0]
  Self-evolving agents improve by repeatedly proposing changes to their own prompts, skills, ormodifications.<n>On Qwen2.5 agents (0.5B-3B) self-evolving at the prompt level on GSM8K, SVAMP, and ARC-Challenge, greedy acceptance commits 30-42% false and 10-33% harmful edits.<n> PACE commits the real one and essentially nothing else, matching greedy's held-out accuracy at sharply lower variance and about 18% lower evaluation cost.
  arXiv  Detail & Related papers  (2026-06-06T11:12:11Z)
• SPOQ: Specialist Orchestrated Queuing for Multi-Agent Software Engineering [2.0769172070951067]
  Multi-agent AI systems show promise for automating software engineering tasks.<n>Existing approaches suffer from coordination overhead, quality control gaps, and limited human oversight.<n>We introduce SPOQ (Specialist Orchestrated Queuing), a methodology combining three innovations.
  arXiv  Detail & Related papers  (2026-06-02T03:59:56Z)
• SOPBench: Evaluating Language Agents at Following Standard Operating Procedures and Constraints [59.645885492637845]
  SOPBench is an evaluation pipeline that transforms each service-specific SOP code program into a directed graph of executable functions.<n>Our approach transforms each service-specific SOP code program into a directed graph of executable functions and requires agents to call these functions based on natural language SOP descriptions.<n>We evaluate 18 leading models, and results show the task is challenging even for top-tier models.
  arXiv  Detail & Related papers  (2025-03-11T17:53:02Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.