1
Decomposing how prompting steers behavior
arXiv论文揭秘提示词如何无参数更新地引导模型行为，揭示背后机制
arXiv:2606.03093v1 Announce Type: new Abstract: Prompting steers large language models (LLMs) and vision-language models (VLMs) without weight updates…
2
Detection Without Correction: A Two-Parameter Decomposition of Multi-Stage LLM Pipelines
无需纠正即可检测，两参数分解多阶段LLM管道，揭示内部机制新视角
arXiv:2605.27559v1 Announce Type: cross Abstract: Multi-stage LLM pipelines that perform multi-agent debate, intrinsic self-correction, or retrieval-a…
3
Probabilistic Attribution For Large Language Models
利用条件概率解析大模型生成逻辑，揭示训练与推理中的分布结构。
arXiv:2605.21726v1 Announce Type: new Abstract: The generative nature of Large Language Models (LLMs) is reflected in the conditional probabilities th…