HABIT-Bench: Evaluating Habit Induction from Longitudinal Weak Evidence in Agent Memory

Researchers introduced HABIT-Bench, a benchmark for evaluating whether LLM agents can induce user habits from longitudinal weak evidence stored in agent memory, addressing the problem that individual observations provide only partial evidence about a habit. The benchmark targets long-term interactions in which users reveal habits through choices and actions that agents must infer across conversations.

In long-term interactions with large language model LLM agents, users reveal habits through their choices and actions. These habits capture implicit user information that agents must infer across conversations. However, individual observations may provide only partial evidence about a habit or whe