After an 80 percent price reduction, GPT-5.6 Luna costs $0.20 per million input tokens and is thus cheaper than Google’s Gemini 3.1 Flash-Lite and one-fifth…
Direct-OPD transfers RL-induced policy shifts from weaker to stronger models by leveraging the implicit reward signal from the log-ratio of the RL-shifted and original policy.
Different layers perform different roles and could therefore enable non-uniform distribution of parameters and computational resources as an alternative to constant architectural width.
Hidden-state alignment reduces sampling variance, closes the student-teacher gap more effectively, and trains with less memory and computational time than output-only distillation.
ThoughtFold identifies and removes redundant exploration steps in reasoning chains, reducing token consumption by 56% for DeepSeek-R1-Distill-Qwen-7B while maintaining state-of-the-art accuracy.
