Unverified80% confidenceFactTime unknown
GPT-OSS采用了与O4模型完全相同的三阶段训练流程：模型预训练、全量指令微调、RL强化学习后训练
1
Sources
80%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe self-training breakthrough in GPT-5.3 extended the self-evaluation mechanism to the entire training data pipeline, meaning the training materials themselves are also generated by the model, not just used for evaluation.75% similarUnverifiedGPT 系列、Llama 等模型的核心能力绝大部分来自预训练而非后续微调阶段73% similarVerifiedA single complete training run of a GPT-4-level large language model can consume tens of gigawatt-hours (GWh) of electricity.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/23210API
curl https://kongchang.com/api/v1/knowledge/claims/23210MCP
get_claim(id=23210)