What question did this study set out to answer?

This research aims to develop and evaluate Reminisce, a memory architecture that mirrors human cognitive processes in AI systems.

May 10, 2026Open Access

Reminisce: A Cognitive Science-Inspired Memory Architecture for AI Agents

Puntos clave

This research aims to develop and evaluate Reminisce, a memory architecture that mirrors human cognitive processes in AI systems.
Developed Reminisce as a three-tier memory architecture (working, episodic, semantic memory).
Evaluated performance on LongMemEvalS using three Claude model tiers with 500 questions.
Measured accuracy and precision across different model selections.
Overall accuracy ranged from 49.4% (Haiku) to 55.8% (Opus) across model tiers.
Precision for attempted answers remained constant at approximately 81% across different tiers.
Model selection influenced coverage of responses but not precision.

Resumen

Long-term memory systems for AI agents predominantly optimize for verbatim retrieval. We present Reminisce, an open-source memory architecture that models the cognitive pipeline of human memory through three distinct tiers: working memory (bounded capacity buffer), episodic memory (timeline of experiences), and semantic memory (consolidated knowledge with contradiction detection). We evaluate on LongMemEvalS (500 questions across 6 question types) using three Claude model tiers. Overall accuracy ranges from 49.4% (Haiku) to 55.8% (Opus), with precision on attempted answers approximately constant at 81% across tiers. We observe that model selection affects coverage but not precision within a fixed retrieval architecture. Reminisce is released as open-source software.

Leer artículo completoexternamente

Connected Papers

Building similarity graph...

Analyzing shared references across papers

Discussion

Institutions

Summit School

References and Citations

Add This Paper to Your Research Feed

Any time a new paper drops it will be there.