Direct Preference Optimization (DPO)

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• Direct Preference Optimizationframework0.886
  Post-training alignment method during which undesirable behaviors emerged in the studied model.
• Deontological optimizationconcept0.742
  Predictive accuracy applies pressure directly on actions rather than consequences, avoiding instrumental convergence.
• Grouped Relative Policy Optimization (GRPO)framework0.722
  Cost-efficient training algorithm used by DeepSeek-R1 for RL-based reasoning
• Deep Optimizationframework0.721
• GRPO (Group Relative Policy Optimization)method0.719
  RL algorithm used to train the activation verbalizer on open models; samples group of candidate descriptions and applies policy optimization.
• DAS finds better alignments than brute-force search by using gradient descent rather than exhaustive discrete searchclaim0.712
  Second central claim of the paper.
• Preference Engineering of Digital Mindsconcept0.707
  The ethical question of whether precision-engineering digital mind preferences to support human incumbents is procedurally permissible
• Acting to optimize value and perception are two aspects of exactly the same principle: minimization of free energy.claim0.707
  Foundational claim unifying action and perception within single optimization framework.