EVAF: A Test-Retest Protocol for Selective Parametric Consolidation
2026-06-29 • Machine Learning
Machine LearningArtificial Intelligence
AI summaryⓘ
The authors developed a method called EVAF to help language models decide which experiences to keep inside their own 'memory' instead of just retrieving them when needed. They tested this with GPT-2 and TinyLlama, showing that EVAF helps models remember surprising or important information better while keeping normal facts accessible. Their approach reduces confusion between different 'personalities' or styles the model might have and makes the memory changes stable even after interference. This work suggests that remembering by recalling facts and truly internalizing experiences are two different processes in language models.
Language modelsMemory consolidationLoRA (Low-Rank Adaptation)GPT-2TinyLlamaRetrieval systemsParametric memoryInterferenceBehavioral persistence
Authors
Haoliang Han
Abstract
Long-running language agents need mechanisms for deciding which experiences should persist after the working context is gone. Retrieval systems can reinsert past text, but they do not by themselves show that an experience has been selectively consolidated into the model's own behavior. We introduce EVAF, an Echo-Valence Attractor Field mechanism for gated LoRA consolidation, and a test-retest protocol for measuring selective parametric consolidation under controlled interference. Across GPT-2 and TinyLlama, EVAF preferentially consolidates high-valence, high-surprise experiences while preserving retrieval-accessible factual memory through a complementary routed memory path. Test-retest measurements show stronger post-interference behavioral persistence than frozen, retrieval-only, and ungated continual-update baselines, while keeping parameter drift and cross-persona contamination low. The results support a separation between memory access and memory depth: retrieving a fact and internalizing an experience are distinct computational operations.