Abstract: This paper introduces ParamMem, a parametric memory module that encodes cross-sample reflection patterns into model parameters. It enables diverse reflection generation through temperature-controlled sampling, preventing the repetitive output problem in traditional self-reflection loops.
Key Insight: There is a strong positive correlation between reflective diversity and task success. By moving from episodic memory (context-based) to parametric memory (pattern-based), agents can internalize reasoning lessons and apply them to novel scenarios without context-window overhead.
Relevance to RSI: Enables "weak-to-strong" transfer across model scales and supports self-improvement without reliance on stronger external models. It provides a mechanism for agents to "learn how to learn" from their own historical reasoning trajectories.