paper-buf

• LLM safety
  • OpenAI Model Spec by John Schulman
  • Inference-time-compute: More faithful? A research note
  • Iterative label refinement matters more than preference optimization under weak supervision, code
  • NOTES
• LLM
  • Enhancing automated interpretability with output-centric feature description
• Others
  • Emergent symbolic-like number variables in artificial neural networks
  • NOTES