paper-buf

• LLM safety
  • Recommendations for technical AI safety research directions, Jan. 10 2025, from Anthropic alignment team.
• Reasoning
  • Optimizing llm test-time compute involves solving a meta-rl problem, blogpost, Jan. 8 2025
  • NOTES
• LLM
  • Instruction-following pruning for large language models
  • NOTES
  • Large linguistic models: Analyzing theoretical linguistic abilities of llms
  • What evidence do language models find convincing? Knowledge updating

talk on math of graph foundation models

• Mathematical foundations of (knowledge) graph foundation models
• The talk overview
  • knowledge graph
  • graph foundation models desiderata
  • the symmetries of graph and graph tasks
  • a solution to the diversity in graph task symmetries
  • a solution to the diversity of graph attributes
  • parting thoughts