# HABIT-Bench: Evaluating Habit Induction from Longitudinal Weak Evidence in Agent Memory > Source: > Published: 2026-10-28 02:51:35+00:00 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