Unverified50% confidenceFactExact time
MiniCPM5-2B开源了训练数据，包括约50万条agent样本和8万多条强化学习样本
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/16/2026
First Seen
Valid until: 12/15/2026
Sources
Related Entities
Related Claims
Unverified开源模型LLaMA 2使用了2万亿token训练数据77% similarUnverified可利用大模型基于领域文档自动生成合成训练数据来构造问答对，通常需要数千至数万对训练样本才能取得显著效果73% similarUnverified后训练阶段使用几万对到几十万对'用户请求+标准回答'的配对数据来训练模型理解人类指令意图73% similarUnverified原始DETR在COCO数据集（约11.8万张训练图像）上需要训练500个epoch才能达到理想性能70% similarUnverifiedMiniMind 覆盖了Tokenizer训练、预训练、监督微调SFT、LoRA微调、DPO强化学习、模型蒸馏等完整环节69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/924463API
curl https://kongchang.com/api/v1/knowledge/claims/924463MCP
get_claim(id=924463)