Unverified50% confidenceFactExact time
应用算法是在已有大模型基础上进行二次开发、微调、本地化部署和私有数据训练，侧重工程实践，典型技术包括LoRA、QLoRA等参数高效微调方法和RAG检索增强生成技术，一本本科基本能够得上
1
Sources
50%
Confidence
Long-term
Relevance
8/18/2026
First Seen
Sources
Related Claims
UnverifiedThe training recipes published by SWE-bench cover key aspects including data preparation, model selection, fine-tuning strategies, and inference frameworks.72% similarUnverified专用SWE模型通常采用领域数据筛选、持续预训练和针对性指令微调三重策略72% similarUnverifiedGPTQ、AWQ等后训练量化算法已成熟，量化部署已成为工业界标配70% similarUnverifiedAn increasing number of real-world scenarios involve calling pre-trained large models, designing prompts, building RAG systems, and embedding models into products.70% similarUnverified通过LoRA等轻量化微调技术，可以用少量特定数据对模型进行定向训练，生成特定类型内容69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/769942API
curl https://kongchang.com/api/v1/knowledge/claims/769942MCP
get_claim(id=769942)