The Teacher Is a Direction, Not a Destination: Extrapolating RL-Induced Representation Residuals in On-Policy Distillation

Open paper

Benchmark	Model	Rank	Results
mathematical-reasoning-on-aime24	RIDE (Qwen3-4B -> Just-Qwen3-4B)	#63	Acc: 63.13
mathematical-reasoning-on-aime-2025	RIDE (Qwen3-4B -> Just-Qwen3-4B)	#81	Accuracy: 52.92