Summary and Comments on Anthropic's Pilot Sabotage Risk Report
"对齐" 相关文章
ImpossibleBench: Measuring Reward Hacking in LLM Coding Agents
RL记得更牢，SFT更健忘？普林斯顿陈丹琦团队改写后训练认知
FLORA: Unsupervised Knowledge Graph Alignment by Fuzzy Logic
⿻ Symbiogenesis vs. Convergent Consequentialism
Will AI superintelligence kill us all? (with Nate Soares)
GTAlign: Game-Theoretic Alignment of LLM Assistants for Mutual Welfare
Realistic Reward Hacking Induces Different and Deeper Misalignment
How to Build an Advanced Voice AI Pipeline with WhisperX for Transcription, Alignment, Analysis, and Export?
Robust Preference Optimization: Aligning Language Models with Noisy Preference Feedback
DeepSeek r1是一个极不安全的 AI 模型，而开源让它失去控制
当AI学会欺骗，我们该如何应对？
Ethics2vec: aligning automatic agents and human preferences
Selective Generalization: Improving Capabilities While Maintaining Alignment
A Comprehensive Survey of Direct Preference Optimization: Datasets, Theories, Variants, and Applications
Advanced fine-tuning methods on Amazon SageMaker AI
Do Self-Perceived Superintelligent LLMs Exhibit Misalignment?
Foom & Doom 2: Technical alignment is hard
Case Studies in Simulators and Agents
Reward button alignment
