ENGRAM: Effective, Lightweight Memory Orchestration for Conversational Agents
About
Large language models (LLMs) deployed in user-facing applications require long-horizon consistency: the ability to remember prior interactions, respect user preferences, and ground reasoning in past events. However, contemporary memory systems often adopt complex architectures such as knowledge graphs, multi-stage retrieval pipelines, and OS-style schedulers, which introduce engineering complexity and reproducibility challenges. We present ENGRAM, a lightweight memory system that organizes conversation into three canonical memory types (episodic, semantic, and procedural) through a single router and retriever. Each user turn is converted into typed memory records with normalized schemas and embeddings and stored in a database. At query time, the system retrieves top-k dense neighbors for each type, merges results with simple set operations, and provides the most relevant evidence as context to the model. ENGRAM attains state-of-the-art results on LoCoMo, a multi-session conversational QA benchmark for long-horizon memory, and exceeds the full-context baseline by 15 points on LongMemEval while using only about 1% of the tokens. These results show that careful memory typing and straightforward dense retrieval can enable effective long-term memory management in language models without requiring complex architectures.
Key insights
What problem does ENGRAM solve for agent memory?
LLM agents need long-horizon memory but current systems use complex graphs or multi-stage schedulers. ENGRAM routes each turn into episodic, semantic, or procedural stores, embeds records in SQLite, and retrieves top-k dense neighbors per type to answer queries.
How does ENGRAM differ from prior memory systems?
Unlike graph-based or scheduler-heavy systems, ENGRAM uses typed separation with simple dense retrieval. On LoCoMo it scores 77.55 overall LLM-Judge versus memOS 72.99 and full-context 72.60, using about 35% fewer memory tokens.
What are the practical uses of these findings?
ENGRAM is a benchmark prototype, not a documented deployment. Its compact, typed retrieval suits applications needing efficient long-horizon memory. The authors suggest tutoring, customer support, and on-device assistants as promising domains.
What limitations and future work do the authors note?
Effectiveness depends on dense retrieval quality; paraphrased facts outside the embedder's neighborhood can cause misses. The minimal router may mishandle complex multi-category utterances. Evaluation is text-only, English-centric, and uses GPT-4o-mini as judge, introducing potential bias.