1
Towards Understanding Steering Strength
ICML 2026论文深入解析模型引导技术中的强度量化问题，为AI可解释性提供新视角。
arXiv:2602.02712v2 Announce Type: replace Abstract: A popular approach to post-training control of large language models (LLMs) is the steering of int…
2
Graph-Regularized Sparse Autoencoders for LLM Safety Steering
提出图正则化稀疏自编码器，提升大模型安全行为干预的精准度。
arXiv:2512.06655v3 Announce Type: replace-cross Abstract: Sparse autoencoders (SAEs) are increasingly used to extract activation directions for infere…
3
You can't whisper at an AI agent
与AI agent对话，清晰指令比窃窃私语更有效。这篇来自Stripe团队的实践洞察，揭示了代理交互中指令粒度与系统响应的微妙关系——你的"呢喃"可能被当作噪声，而明确意图才是驱动智能的关键。
Article URL: https://stripe.dev/blog/ai-steering-experiments Comments URL: https://news.ycombinator.com/item?id=48162696 Points: 1 # Comments: 1