Research Papers 论文研究 5h ago Updated 1h ago 更新于 1小时前 46

The Halt Vector: Internalizing a Causal Steering Intervention for Efficient Reasoning 停止向量:将因果干预内部化以实现高效推理

The paper introduces a "halt vector" at layer 18 of DeepSeek-R1-Distill-Qwen-7B that controls reasoning length by internalizing a causal steering intervention directly into model weights The chain of thought runs approximately twice as long as the model's answer probability settles, with removable excess varying per problem A naive approach of maximizing scalar projection onto the halt direction corrupts off-axis dimensions and paradoxically lengthens generation; the solution requires reconstruc 推理模型在已知答案后仍会过度思考,思考链长度约为答案概率稳定所需长度的约两倍 提出"halt vector"机制,通过模型第18层的差异均值方向控制思考时长,而非复制值轴 将因果干预内化到权重中需重建整个受控激活,同时固定非轴维度为自然值,避免破坏下游读取器 仅从24个问题拟合且无需强化学习,在5个未见基准测试中保持准确率的同时减少约25%思考量 该方法有效关闭了随问题难度增长的非终止病理问题

60
Hot 热度
75
Quality 质量
65
Impact 影响力

Analysis 深度分析

TL;DR

  • The paper introduces a "halt vector" at layer 18 of DeepSeek-R1-Distill-Qwen-7B that controls reasoning length by internalizing a causal steering intervention directly into model weights
  • The chain of thought runs approximately twice as long as the model's answer probability settles, with removable excess varying per problem
  • A naive approach of maximizing scalar projection onto the halt direction corrupts off-axis dimensions and paradoxically lengthens generation; the solution requires reconstructing steered activations while pinning off-axis dimensions to natural values
  • The method removes ~25% of thinking tokens at held accuracy across five unseen benchmarks, with per-problem slack tracking at 0.70 correlation
  • The halt vector also eliminates a non-termination pathology that worsens with difficulty and is exacerbated by decoding-time confidence hooks

Why It Matters

This work bridges causal interpretability and efficient reasoning by showing that steering interventions can be internalized into model weights rather than applied only at inference time, offering a fundamentally different approach to controlling reasoning length. For AI practitioners working with reasoning models, it demonstrates that post-hoc length penalties and decoding-time early exits have limitations that weight-level interventions can address, particularly for problems with variable removable slack.

Technical Details

  • Halt Vector Identification: A difference-of-means direction at layer 18 of DeepSeek-R1-Distill-Qwen-7B, identified through causal interpretability analysis, where steering strength directly controls reasoning duration while a replicated value axis produces no effect
  • Weight Installation Challenge: Simply maximizing scalar projection onto the halt direction corrupts off-axis dimensions that downstream frozen readers depend on, causing generation to lengthen instead of shorten; the working approach reconstructs the full steered activation while constraining off-axis dimensions to their natural values
  • Training Setup: Fit from only 24 problems with no reinforcement learning, demonstrating data efficiency
  • Evaluation: Tested across five unseen benchmarks, achieving ~25% token reduction at held accuracy with per-problem removable slack tracking correlation of 0.70
  • Pathology Closure: The intervention eliminates a non-termination behavior that scales with problem difficulty, which decoding-time confidence-based hooks actually worsen

Industry Insight

  • Weight-level causal steering represents a promising direction for making reasoning models more efficient without sacrificing accuracy, particularly as reasoning models continue to grow in length and cost
  • The finding that decoding-time confidence hooks exacerbate non-termination suggests practitioners should be cautious about naive confidence-based early stopping in reasoning models
  • The 0.70 correlation between the halt's cut and each problem's removable slack indicates the approach generalizes across problem types, making it a viable strategy for production deployments where reasoning overhead is a concern

TL;DR

  • 推理模型在已知答案后仍会过度思考,思考链长度约为答案概率稳定所需长度的约两倍
  • 提出"halt vector"机制,通过模型第18层的差异均值方向控制思考时长,而非复制值轴
  • 将因果干预内化到权重中需重建整个受控激活,同时固定非轴维度为自然值,避免破坏下游读取器
  • 仅从24个问题拟合且无需强化学习,在5个未见基准测试中保持准确率的同时减少约25%思考量
  • 该方法有效关闭了随问题难度增长的非终止病理问题

为什么值得看

这项工作展示了如何将可解释性发现转化为实际的模型优化手段,为推理模型的效率改进提供了新的技术路径。对于需要长链思考的推理模型,这种基于内部机制的干预方法比传统的长度惩罚或解码时早期退出更具针对性。

技术解析

  • Halt Vector机制:在DeepSeek-R1-Distill-Qwen-7B模型的第18层识别出一个差异均值方向,该方向的 steering strength 可控制模型思考时长,而复制值轴方向无效
  • 权重安装方法:直接最大化标量投影会破坏下游读取器依赖的非轴维度,导致生成反而更长;正确方法是在重建受控激活时将这些维度固定为自然值
  • 实验设置:仅使用24个问题拟合,无需强化学习,在5个未见基准测试上评估
  • 性能表现:在保持准确率的前提下减少约四分之一的思考量,且削减量与每个问题的可移除冗余度相关性达0.70
  • 病理修复:解决了推理模型中随难度增长的非终止问题,而解码时的置信度钩子会加剧此问题

行业启示

  • 可解释性研究可直接转化为模型优化手段,因果干预的内化比外部解码策略更具潜力
  • 推理模型的"思考效率"优化应从内部机制入手,而非依赖全局长度惩罚等粗粒度方法
  • 对于部署场景,这种无需额外强化学习训练的轻量级干预方法更具实用价值

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。

Research 科学研究 LLM 大模型 Inference 推理 Alignment 对齐