Reward-Aligned On-Policy Distillation
RA-OPD improves model performance on math and code tasks, making it valuable for developers working with LLMs in technical domains.
RA-OPD improves model performance on math and code tasks, making it valuable for developers working with LLMs in technical domains.