PEARL: Self-Evolving Assistant for Time Management with RL

ACL ARR 2026 January Submission1689 Authors

Keywords: AI Agent, LLMs, Preference Alignment, Evaluation

Abstract: Overlapping calendar invitations force busy professionals to repeatedly decide which meetings to attend, reschedule, or decline. We refer to this preference-driven decision process as calendar conflict resolution. Automating such process is crucial yet challenging: Scheduling logistics drain hours, and human delegation often fails at scale, raising the question of whether large language model (LLM) agents can reliably learn and apply user preferences to manage time. To enable systematic study, we introduce CalConflictBench, a benchmark for long-horizon calendar conflict resolution. Conflicts are presented sequentially and agents receive feedback after each round, requiring them to infer and adapt to user preferences progressively. Our experiments show that current LLM agents perform poorly with high error rates, e.g., Qwen-3-30B-Think with 0.35 average error rate. To address this gap, we propose PEARL, a reinforcement-learning framework that augments language agent with an external memory module and optimized round-wise reward design, enabling agent to progressively infer and adapt to user preferences on-the-fly. Experiments on CalConflictBench shows that PEARL achieves 0.76 error reduction rate, and 55% improvement on average error rate compared to the strongest baseline.

Paper Type: Long

Research Area: AI/LLM Agents

Research Area Keywords: Autonomous agents, LLM agents, agent evaluation

Contribution Types: Model analysis & interpretability, NLP engineering experiment, Data resources

Languages Studied: English

Submission Number: 1689