Expired60% confidenceFactExact time
HY3 introduces improved anti-hallucination techniques, more reliable tool-calling capabilities, and employs higher-quality post-training data and an enhanced reinforcement learning strategy.
2
Sources
60%
Confidence
Medium-term (~90 days)
Relevance
7/13/2026
First Seen
Valid until: 10/11/2026(expired)
Sources
Related Claims
Unverifiedfal used a new dataset to post-train H3 Max, with a focus on reinforcing prompt-following capabilities and visual quality performance.73% similarUnverifiedSWE-2 is post-trained on Kimi K3 and uses reinforcement learning (RL), optimizing along both the cost and capability dimensions simultaneously.68% similarUnverified后训练优化包括指令微调、RLHF或DPO策略调整、系统提示词工程改进及工具调用和多模态融合能力强化68% similarUnverifiedEMODE uses three mechanisms to support training: Orthogonal Expert Guidance (OEG), Semantic-Acoustic Alignment (SAA), and Gated Diversity Regularization (GDR).68% similarUnverifiedo3-mini-high has been trained with reinforcement learning for reasoning, achieving reasoning capabilities close to o1 at a smaller scale.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/495165API
curl https://kongchang.com/api/v1/knowledge/claims/495165MCP
get_claim(id=495165)