RLHF: A comprehensive Survey for Cultural, Multimodal and Low Latency Alignment Methods
"LLM对齐" 相关文章
Control Barrier Function for Aligning Large Language Models
Meta-Learning Objectives for Preference Optimization
The Sign Estimator: LLM Alignment in the Face of Choice Heterogeneity
Self-Rewarding PPO: Aligning Large Language Models with Demonstrations Only
Users as Annotators: LLM Preference Learning from Comparison Mode
DeAL: Decoding-time Alignment for Large Language Models
H3Fusion: Helpful, Harmless, Honest Fusion of Aligned LLMs
Reward Model Routing in Alignment
UniAPL: A Unified Adversarial Preference Learning Framework for Instruct-Following
Clean First, Align Later: Benchmarking Preference Data Cleaning for Reliable LLM Alignment
Targeting Alignment: Extracting Safety Classifiers of Aligned LLMs
Beyond Ordinal Preferences: Why Alignment Needs Cardinal Human Feedback
"Pull or Not to Pull?'': Investigating Moral Biases in Leading Large Language Models Across Ethical Dilemmas
PROPS: Progressively Private Self-alignment of Large Language Models
Aligning LLMs on a Budget: Inference-Time Alignment with Heuristic Reward Models
Alignment and Safety in Large Language Models: Safety Mechanisms, Training Paradigms, and Emerging Challenges
Inverse Reinforcement Learning Meets Large Language Model Post-Training: Basics, Advances, and Opportunities
Crome: Google DeepMind’s Causal Framework for Robust Reward Modeling in LLM Alignment
I replicated the Anthropic alignment faking experiment on other models, and they didn't fake alignment
