The Price of Meaning: Why Every Semantic Memory System Forgets

arXiv cs.AI / 3/31/2026

💬 OpinionSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper proves that semantic memory systems—where retrieval is organized by meaning in an embedding space—inevitably face interference, forgetting, and false recall due to the underlying geometry that enables semantic generalization.
It formalizes a tradeoff for semantically continuous kernel-threshold memories and derives results including finite effective rank for useful representations, inevitable “competitor” mass in retrieval neighborhoods, and power-law forgetting under certain arrival statistics.
It shows that false recall cannot be fully removed by threshold tuning for associative lures that satisfy a delta-convexity condition.
Experiments across five architectures (vector, graph, attention-based context, BM25 retrieval, and parametric memory) support the theory: pure semantic systems show direct forgetting/false recall, while reasoning-augmented systems may shift degradation into more severe failure modes.
The authors conclude that avoiding interference requires sacrificing semantic generalization, so “the price of meaning” is unavoidable across tested architectures.

Abstract

Every major AI memory system in production today organises information by meaning. That organisation enables generalisation, analogy, and conceptual retrieval -- but it comes at a price. We prove that the same geometric structure enabling semantic generalisation makes interference, forgetting, and false recall inescapable. We formalise this tradeoff for \textit{semantically continuous kernel-threshold memories}: systems whose retrieval score is a monotone function of an inner product in a semantic feature space with finite local intrinsic dimension. Within this class we derive four results: (1) semantically useful representations have finite effective rank; (2) finite local dimension implies positive competitor mass in retrieval neighbourhoods; (3) under growing memory, retention decays to zero, yielding power-law forgetting curves under power-law arrival statistics; (4) for associative lures satisfying a

\delta

-convexity condition, false recall cannot be eliminated by threshold tuning. We test these predictions across five architectures: vector retrieval, graph memory, attention-based context, BM25 filesystem retrieval, and parametric memory. Pure semantic systems express the vulnerability directly as forgetting and false recall. Reasoning-augmented systems partially override these symptoms but convert graceful degradation into catastrophic failure. Systems that escape interference entirely do so by sacrificing semantic generalisation. The price of meaning is interference, and no architecture we tested avoids paying it.