OneDayAgent: Towards a Long-Horizon Harness for Autonomous Agents

2026-08-04

AI Summary

This work addresses the compounded challenges of goal drift, state loss, and context overflow that large language model (LLM) agents encounter in long-horizon, cross-environment, multimodal everyday tasks. To this end, we propose the first general-purpose long-horizon agent framework, which decomposes complex tasks into bounded subtasks, employs execution-aware memory management to alleviate context window pressure, and integrates output validation and repair mechanisms to ensure result reliability. The framework natively supports multimodal inputs and heterogeneous tool invocation, and achieves stable generalization across five LLMs from three distinct model families without requiring backend-specific tuning. Evaluated on the AgentIF-OneDay benchmark comprising 104 tasks, our approach attains a state-of-the-art overall score of 0.821 when powered by GLM-5.2.

Abstract

LLM agents are increasingly applied to open-ended everyday requests that span work, study, and life. These tasks are long-horizon, cross-environment, and multimodal, forcing the agent to preserve goals and constraints across many steps while navigating heterogeneous tools and attachments. While prior work has addressed individual failure modes such as goals drift, states loss, and context overflow, whether a single harness can manage them jointly and remain effective across backends has received less study. We present OneDayAgent, a long-horizon harness for autonomous agents. OneDayAgent turns an open-ended request into a managed execution process that decomposes tasks into bounded subtasks, maintains execution memory under context pressure, and verifies and repairs the final deliverable. We evaluate OneDayAgent on AgentIF-OneDay across 104 tasks. With the GLM-5.2 backend, OneDayAgent sets a new state of the art with an overall score of 0.821. The same harness runs across five backend LLMs from three model families, indicating the harness generalizes across backends without tuning, even as different models induce distinct execution styles under the same workflow.

Problem

• long-horizon tasks
• autonomous agents
• goal preservation
• cross-backend generalization
• execution robustness

Innovation

• long-horizon planning
• autonomous agents
• execution memory
• cross-backend generalization
• task decomposition