Tag: reasoning
All the articles with the tag "reasoning".
RLVR 与 GRPO:让小模型拥有推理能力
2026 年的突破不在参数规模,而在后训练。通过 RLVR、GRPO 与 On-Policy DPO,小推理模型已能媲美专有系统。
All the articles with the tag "reasoning".
2026 年的突破不在参数规模,而在后训练。通过 RLVR、GRPO 与 On-Policy DPO,小推理模型已能媲美专有系统。