Unverified60% confidenceFactExact time
o3-mini-high has been trained with reinforcement learning for reasoning, achieving reasoning capabilities close to o1 at a smaller scale.
2
Sources
60%
Confidence
Long-term
Relevance
2/5/2025
First Seen
Sources
Related Entities
Related Claims
VerifiedOpenAI's o-series (o1, o3) are specifically reinforcement-trained for chain-of-thought reasoning capabilities, incorporating Monte Carlo Tree Search (MCTS) and process reward models.71% similarVerifiedUsing the Adam optimizer with a default learning rate of 1e-3 as a starting point is a robust entry strategy for deep learning.67% similarUnverified小模型通过知识蒸馏通常比直接在标注数据上训练精度高出2-5个百分点67% similarUnverifiedA contemporaneous study by O'Brien et al. found that when mid-training was used to teach the model that misaligned behavior is acceptable within a specified context (marked with the invented token <quarantine_token>), and SFT/RL training data conforming to that format was used, misalignment generalization could be reduced, but not beyond the effect of inoculation prompting.66% similarUnverifiedOmniVerifier-C1 adopts a decoupled reinforcement learning training strategy, splitting the optimization objective into two parts: sub-question-level binary judgment and global alignment scoring.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/656API
curl https://kongchang.com/api/v1/knowledge/claims/656MCP
get_claim(id=656)