Interaction as Intelligence Part II: Asynchronous Human-Agent Rollout for Long-Horizon Task Training

TL;DR

This paper introduces Apollo, a novel human-in-the-loop sampling framework that enhances training of large language model agents on long-horizon, domain-specific tasks by integrating asynchronous human guidance and data filtering, leading to significant performance improvements.

Contribution

Apollo's lightweight, asynchronous human guidance approach enables efficient long-horizon task training with reduced annotation costs and improved data quality, addressing limitations of existing methods.

Findings

Apollo achieves over 50% improvement on InnovatorBench with GLM-4.5.

It sustains over 30 hours of human-agent interaction.

It outperforms variants without human interaction by 28%.

Abstract

Large Language Model (LLM) agents have recently shown strong potential in domains such as automated coding, deep research, and graphical user interface manipulation. However, training them to succeed on long-horizon, domain-specialized tasks remains challenging. Current methods primarily fall into two categories. The first relies on dense human annotations through behavior cloning, which is prohibitively expensive for long-horizon tasks that can take days or months. The second depends on outcome-driven sampling, which often collapses due to the rarity of valid positive trajectories on domain-specialized tasks. We introduce Apollo, a sampling framework that integrates asynchronous human guidance with action-level data filtering. Instead of requiring annotators to shadow every step, Apollo allows them to intervene only when the agent drifts from a promising trajectory, by providing prior knowledge, strategic advice, etc. This lightweight design makes it possible to sustain interactions for over 30 hours and produces valuable trajectories at a lower cost. Apollo then applies supervision control to filter out sub-optimal actions and prevent error propagation. Together, these components enable reliable and effective data collection in long-horizon environments. To demonstrate the effectiveness of Apollo, we evaluate it using InnovatorBench. Our experiments show that when applied to train the GLM-4.5 model on InnovatorBench, Apollo achieves more than a 50% improvement over the untrained baseline and a 28% improvement over a variant trained without human interaction. These results highlight the critical role of human-in-the-loop sampling and the robustness of Apollo's design in handling long-horizon, domain-specialized tasks.