Memory canonicalization improves emotional consistency across language models
Memory Canonicalization: A Framework and Benchmark for Cross-Model Drift in Persistent LLM Memory
Artificial Intelligence
Summary
Different large language models (LLMs) can interpret the same stored memory differently, especially in emotional tone or factual meaning. The authors suggest a way to rewrite these memories into a clear, standardized form that separates emotion explicitly. They tested this method using multiple models and found it can improve emotional consistency, though results were not strong enough to be definitive. Their work provides a foundation for improving shared memory use across different AI models in the future.
What this means in practice
- •For llm system developers: Improve consistency of stored memories interpreted by different LLMs by using canonicalized memory representations.
- •For chatbot designers: Enhance emotional tone accuracy in multi-model chatbot systems by structuring memory with explicit emotional valence.
Authors
Amit Vadnere, Aishwarya Lonarkar
Abstract
Persistent memory for Large Language Models (LLMs) has matured rapidly: systems such as MemGPT/Letta, Mem0, and Zep now provide agents with tiered, temporally-aware, model-agnostic external storage, while the Model Context Protocol (MCP) standardizes access to memory servers. A less addressed problem is that an identical stored memory object, retrieved by two different LLMs under otherwise identical conditions, may not be interpreted the same way, factually or emotionally. This paper proposes memory canonicalization: a write-time pipeline that detects ambiguity, conditional structure, and emotional loading in a raw memory object and rewrites it into an explicit, structurally disambiguated canonical form, with emotional valence represented as a separate field rather than inferred from tone. We formalize the pipeline, define a companion Cross-Model Semantic Drift / Emotional Consistency Score benchmark (CMSC-E), and report results from a three-arm pilot using 176 synthetic memory objects and three downstream model families. We find an uncorrected improvement in cross-model emotional consistency for fully canonicalized memory relative to raw memory (+0.050, 95% bootstrap CI [0.013, 0.086], paired t-test p = 0.010), but this result does not survive Bonferroni, Holm, or Benjamini-Hochberg correction across the six comparisons tested. None of the factual-drift (CMSD) comparisons reach significance at any correction level. We report these results as exploratory rather than confirmatory and outline needed follow-up work, including larger samples, independent judge models, human-validated rendering, and preregistration.