Comment on Does RL Incentivize Reasoning in LLMs Beyond the Base Model?Comments−kk581yReasoning models aren't really reasoners, its basically neural style transfer protocol where you force a model "decoder" to emit tokens in a style that appears to be Reasoning like a deductive thinking.
Comments
Reasoning models aren't really reasoners, its basically neural style transfer protocol where you force a model "decoder" to emit tokens in a style that appears to be Reasoning like a deductive thinking.