Research Papers 论文研究 16h ago Updated 2h ago 更新于 2小时前 49

Moir: Let the Model Direct Its Own Story for Robust Cross-Domain Knowledge Editing Moir:让模型引导其自身故事以实现稳健的跨领域知识编辑

Moir addresses the asymmetric degradation of core capabilities (math/programming) during knowledge editing by aligning preservation distributions with the model's operative distribution. The method estimates preservation covariance directly from the model’s own decoding distribution using random vocabulary token seeding, bypassing instruction-following templates. Moir requires no external data and serves as a drop-in component for covariance-based editors like MEMIT and AlphaEdit. Experiments on 提出Moir方法,通过从模型自身解码分布采样来估计保留协方差,解决知识编辑中的能力退化问题。 指出传统基于外部语料库(如Wikipedia)的编辑器存在分布不匹配,导致数学和编程推理等核心能力崩溃。 Moir无需外部数据,可作为即插即用组件集成到MEMIT和AlphaEdit等现有编辑器中。 在OLMo-2、Llama-3.1和Qwen-3模型上验证,显著提升了大规模编辑后的能力保持率,如Qwen3-8B在2万次编辑后GSM8K准确率从10.9%提升至79.9%。

65
Hot 热度
75
Quality 质量
70
Impact 影响力

Analysis 深度分析

TL;DR

  • Moir addresses the asymmetric degradation of core capabilities (math/programming) during knowledge editing by aligning preservation distributions with the model's operative distribution.
  • The method estimates preservation covariance directly from the model’s own decoding distribution using random vocabulary token seeding, bypassing instruction-following templates.
  • Moir requires no external data and serves as a drop-in component for covariance-based editors like MEMIT and AlphaEdit.
  • Experiments on OLMo-2, Llama-3.1, and Qwen-3 show significant improvements in retaining reasoning abilities, with Qwen-3-8B maintaining 79.9% GSM8K accuracy after 20,000 edits compared to 10.9% with Wikipedia baselines.

Why It Matters

This research solves a critical bottleneck in deploying knowledge editing systems: the collapse of complex reasoning skills while factual recall remains intact. By demonstrating that models can self-supervise their own preservation distributions, it offers a scalable, data-efficient solution for maintaining model utility during continuous updates. This is particularly relevant for practitioners managing large-scale, iterative knowledge updates where access to original training corpora is restricted.

Technical Details

  • Problem Identification: Asymmetric capability erosion is traced to a distributional mismatch where static external corpora fail to capture the shifted manifold created by post-training steps like SFT and DPO.
  • Methodology: Moir estimates the preservation covariance $C$ by sampling from the model's internal decoding distribution. Seeding generation with a single random vocabulary token exposes broader internalized subspaces without being dominated by instruction-following templates.
  • Implementation: It acts as a modular drop-in component for existing covariance-based editors (MEMIT, AlphaEdit), requiring no additional external datasets or access to proprietary pre/post-training corpora.
  • Evaluation: Tested across OLMo-2, Llama-3.1, and Qwen-3 (7-8B) models in both batch and sequential editing regimes, focusing on encyclopedic recall and mathematical/programmatic reasoning benchmarks like GSM8K.

Industry Insight

  • Self-Supervised Preservation: Organizations should prioritize methods that leverage the model's internal state for preservation constraints, reducing dependency on hard-to-obtain or non-existent training data.
  • Robustness in High-Volume Editing: For systems requiring frequent knowledge updates, integrating Moir-like mechanisms is essential to prevent the rapid degradation of reasoning capabilities, which are often more valuable than simple factual recall.
  • Standardization of Editing Metrics: The stark contrast between baseline and Moir performance highlights the need for rigorous evaluation of reasoning retention, not just factual accuracy, when assessing knowledge editing tools.

TL;DR

  • 提出Moir方法,通过从模型自身解码分布采样来估计保留协方差,解决知识编辑中的能力退化问题。
  • 指出传统基于外部语料库(如Wikipedia)的编辑器存在分布不匹配,导致数学和编程推理等核心能力崩溃。
  • Moir无需外部数据,可作为即插即用组件集成到MEMIT和AlphaEdit等现有编辑器中。
  • 在OLMo-2、Llama-3.1和Qwen-3模型上验证,显著提升了大规模编辑后的能力保持率,如Qwen3-8B在2万次编辑后GSM8K准确率从10.9%提升至79.9%。

为什么值得看

这篇文章揭示了当前知识编辑技术在部署时的主要瓶颈——非对称的能力退化,并提供了无需额外数据的解决方案。对于AI从业者而言,理解如何对齐模型的“操作分布”与“保留分布”是实现非破坏性编辑的关键,这为构建更鲁棒的持续学习系统提供了重要思路。

技术解析

  • 问题诊断:现有的协方差基编辑器仅保留参考语料库张成的子空间,无法捕捉由SFT和DPO等后训练步骤塑造的操作分布,导致静态外部语料库失效。
  • 核心机制:Moir直接从模型自身采样以估计保留协方差$C$。通过使用单个随机词汇表令牌作为生成种子,绕过指令跟随模板,从而暴露模型内部化的更广泛子空间。
  • 通用性与实用性:该方法不需要任何外部数据,且设计为即插即用组件,可适配任何协方差基编辑器,解决了现代LLM预训练和后训练语料库通常不可公开访问的问题。
  • 实验验证:在OLMo-2、Llama-3.1和Qwen-3 (7-8B)上进行测试,涵盖批处理和顺序编辑模式。结果显示Moir在脆弱领域(如数学推理)的保留效果显著优于基线。

行业启示

  • 自监督数据的重要性:在知识编辑中,利用模型自身的内部状态而非依赖外部静态数据可能是更有效的策略,特别是在外部数据不可得或分布偏移严重时。
  • 关注非对称退化:开发者应警惕知识编辑对推理能力(数学、代码)的潜在破坏,而不仅仅关注事实性知识的更新,需采用更全面的评估指标。
  • 模块化编辑架构:将分布估计模块(如Moir)标准化为即插即用组件,有助于降低不同编辑算法的集成成本,加速鲁棒编辑技术的落地应用。

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

LLM 大模型 Research 科学研究 Fine-tuning 微调 Evaluation 评测