Research Papers 论文研究 5h ago Updated 58m ago 更新于 58分钟前 49

Shared circuits predict whether LLMs generalize across formats in arithmetic reasoning 共享电路预测LLM在算术推理中跨格式泛化能力

LLMs exhibit brittleness when generalizing from numeric arithmetic (e.g., "2+5") to verbal formats (e.g., "two plus five"), despite near-perfect numeric performance Attribution patching is used to independently localize the circuits recruited by models for numeric versus verbal arithmetic reasoning across English, Spanish, and Italian Circuit overlap between a model's numeric and verbal circuits predicts its generalization performance across verbal formats without requiring any labeled data The 使用attribution patching技术独立定位LLM解决数值算术(2+5)与文字算术(如"two plus five")问题时的内部电路 电路重叠程度可预测模型跨格式泛化能力,在三个层面得到验证:解释不同文字格式的相对难度、预测泛化最佳情况、识别正确解答的项目 实验覆盖英语、西班牙语、意大利语三种语言,证明方法具有跨语言泛化性 无需任何标注数据即可实现预测,性能与监督探针相媲美

65
Hot 热度
75
Quality 质量
70
Impact 影响力

Analysis 深度分析

TL;DR

  • LLMs exhibit brittleness when generalizing from numeric arithmetic (e.g., "2+5") to verbal formats (e.g., "two plus five"), despite near-perfect numeric performance
  • Attribution patching is used to independently localize the circuits recruited by models for numeric versus verbal arithmetic reasoning across English, Spanish, and Italian
  • Circuit overlap between a model's numeric and verbal circuits predicts its generalization performance across verbal formats without requiring any labeled data
  • The circuit overlap metric accounts for relative difficulty across verbal formats, identifies which formats models generalize to best, and predicts item-level correctness, rivaling supervised probes

Why It Matters

This work provides a mechanistic, interpretable lens into a persistent weakness of LLMs: poor cross-format generalization in reasoning tasks. By showing that internal circuit overlap—not just surface-level features—predicts generalization ability, it offers practitioners a data-efficient diagnostic tool for evaluating model robustness without expensive labeled benchmarks.

Technical Details

  • Method: Attribution patching is employed to localize the computational circuits each model recruits for solving numeric arithmetic problems (e.g., "2+5") versus verbal renditions (e.g., "two plus five," "dos más cinco," "due più cinque")
  • Multilingual scope: The study evaluates generalization across three languages—English, Spanish, and Italian—allowing cross-linguistic comparison of circuit overlap and its predictive power
  • Three-level prediction: Circuit overlap predicts (1) the relative difficulty ranking of the three verbal formats, (2) which verbal formats each model generalizes to best, and (3) which individual items are solved correctly
  • Unsupervised advantage: The approach requires no labeled data and achieves predictive performance rivaling supervised probes, making it a cost-effective diagnostic for model internals

Industry Insight

  • Model evaluation pipelines should incorporate mechanistic interpretability tools like attribution patching to diagnose generalization brittleness before deployment, rather than relying solely on benchmark accuracy
  • The finding that circuit overlap predicts cross-format performance suggests that training strategies aimed at strengthening shared representational circuits could improve robustness to prompt variations—a key concern for real-world LLM applications
  • The unsupervised nature of this diagnostic makes it scalable for continuous model monitoring across languages and domains without the overhead of creating labeled evaluation sets

TL;DR

  • 使用attribution patching技术独立定位LLM解决数值算术(2+5)与文字算术(如"two plus five")问题时的内部电路
  • 电路重叠程度可预测模型跨格式泛化能力,在三个层面得到验证:解释不同文字格式的相对难度、预测泛化最佳情况、识别正确解答的项目
  • 实验覆盖英语、西班牙语、意大利语三种语言,证明方法具有跨语言泛化性
  • 无需任何标注数据即可实现预测,性能与监督探针相媲美

为什么值得看

该研究揭示了LLM在算术推理中跨格式泛化能力的内在机制,为理解模型"脆弱性"来源提供了可解释的视角。对于AI从业者而言,这为提升模型鲁棒性、减少表面形式敏感性问题提供了新的技术路径和评估手段。

技术解析

  • 采用attribution patching技术,分别定位模型处理数值格式和文字格式算术问题时的独立电路
  • 实验涉及三种语言环境:英语("two plus five")、西班牙语("dos más cinco")、意大利语("due più cinque")
  • 核心假设:模型自身数值电路与文字电路的重叠程度,可预测其在文字格式上的表现
  • 验证层面包括:三种文字格式的相对难度排序、模型泛化最佳的条件、具体题目级别的正确/错误预测
  • 方法优势:完全无监督,无需标注数据,性能达到监督探针水平

行业启示

  • 为LLM泛化能力提升提供了可操作的诊断工具:通过分析内部电路重叠度,可提前识别模型在特定格式下的薄弱环节
  • 揭示了当前LLM"表面形式敏感"问题的本质原因,为模型架构优化和训练策略改进指明方向
  • 无监督电路分析方法可推广至其他推理任务,为可解释AI研究提供了新的技术范式

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 Evaluation 评测 Benchmark 基准测试