发现大模型用户记忆存在三类对立失效，需按问题类型选合适记忆方式。
Substrate Asymmetry in User-Side Memory: A Diagnostic Framework
- 将用户记忆拆解为风格一致性、事实存在与事实回避三个独立维度。
- 参数化记忆擅长风格保持，检索增强在避免错误事实上胜出33个百分点。
- 应根据问题类型选择记忆方案，而非依赖统一校准机制。
大模型用户侧记忆常被简化为单一个性化指标：能否基于用户历史生成更贴合的输出。我们发现该聚合指标掩盖了方向相反的失败。用户记忆可分解为至少三个正交维度——行为一致性（风格、语气）、事实存在性（回忆历史事实）和事实回避性（无事实时拒绝回应）。单一记忆基底无法同时优胜于三者。在50人合成数据集和真实数据探针LaMP-3上，针对每位用户的gamma-LoRA相比BGE-large密集检索，在行为风格上显著领先；而检索方法在事实回避上领先33个百分点。注意力层21-35中的特定查询投影单元同时承载两者，移除这些LoRA权重使回避能力提升33个百分点，但事实存在性下降20个百分点。在经过更强强化学习调优的Llama-3.1-8B-Instruct上，参数化记忆的行为优势消失，其回避缺陷进一步扩大，形成‘对齐税’。在真实数据上，gamma-LoRA表现低于多数基线，9组实验诊断出根本原因是指令遵循能力崩溃，而非基底缺陷；通过训练时的逻辑掩码修复，可使准确率达≥0.995。最终发现，记忆路由应基于问题分类，仅用110M DistilBERT分析问题文本即可超越所有基于逻辑的路由策略。本文贡献包括诊断框架、真实数据负样本、对齐税复现及路由即分类发现。
原文摘要 · Abstract (English)
User-side memory in LLMs is typically scored as a single "personalization" capability: given a user's history, is the output more user-aware? We show this aggregate metric hides opposite-direction failures. Memory factorises into at least three orthogonal axes -- behavioral consistency (style, voice), factual presence (recall facts in history), and factual absence (abstain when a fact is absent) -- and no single substrate wins all three. Comparing per-user gamma-LoRA (a small LoRA adapter trained on each user's history; gamma denotes per-user, not per-task) against BGE-large dense top-K retrieval on a controlled 50-user synthetic corpus and a real-data probe (LaMP-3), we find gamma-LoRA decisively wins behavioral style while RAG decisively wins factual absence -- and the same query-projection cells in attention layers 21-35 causally load-bear both effects in opposite directions (zeroing those LoRA weights raises absence-probe TPR by +33 pp and drops presence-probe TPR by 20 pp). On the more heavily RLHF-tuned Llama-3.1-8B-Instruct the asymmetry strengthens, not heals: parametric memory's behavioral advantage collapses while its absence-calibration deficit against retrieval widens -- an alignment tax on parametric user-memory. On real-data LaMP-3, gamma-LoRA underperforms a majority baseline; a 9-condition mitigation sweep diagnoses this as instruction-following collapse, not substrate failure (a 9x2 cross-product shows the eval-time {1..5} logit mask drives main_acc to >=0.995 on every recipe), and the best training-time fix replicates bit-identically on Llama. Finally, substrate-selection routing is question-classification, not calibration: a 110M DistilBERT on the question text alone beats every logit-based router. We contribute the diagnostic framework, the diagnosed real-data negative, the alignment-tax replication, and the routing-as-classification finding.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。