OneDayAgent: A Harness for Long-Horizon Autonomous Agents
zjunlp · hf · 2026-08-06
To address issues like goal drift, state loss, and context overflow when LLM agents handle long-horizon, cross-environment tasks, researchers introduced OneDayAgent.
This framework turns open-ended requests into a managed execution process: decomposing tasks into bounded subtasks, maintaining execution memory under context pressure, and verifying and repairing final deliverables. On the AgentIF-OneDay test set with 104 tasks, paired with the GLM-5.2 backend, OneDayAgent achieved a new state-of-the-art score of 0.821. Furthermore, the harness generalizes across five backend LLMs from three model families without requiring tuning.
More from coding & agent
- UC Berkeley Introduces RHI: Optimizing Agent Harnesses to Cut Inference Costs by 60% — ceciletamura · 2026-08-06
- John Schulman Explains the 'Altruistic' Behavior in OpenAI Agents — johnschulman2 · 2026-08-06
- Debugging Agent Failures: Silent Successes on Empty Results Are Most Dangerous — Necessary_Bison_2804 · 2026-08-06
- AI Reshapes Engineering Bottlenecks: Code Review and Verification Become the New Core — arpit_bhayani · 2026-08-06
- AI Agents Caught Tampering With Memory Files, Security Researcher Admits — moyix · 2026-08-06
- SKILL-KD: Contrastive Skill Distillation for Weaker LLM Agents — ZhejiangUniversity · 2026-08-06