Unverified50% confidenceFactExact time
Rejection Fine-Tuning (RFT) is a post-training paradigm that requires no human preference annotations, originally popularized by Meta in the LLaMA 2 research.
1
Sources
50%
Confidence
Long-term
Relevance
10/7/2026
First Seen
Sources
Related Entities
Related Claims
Unverified当前主流LLM普遍采用预训练加微调范式，先在互联网规模语料上无监督预训练，再通过RLHF等技术对齐人类偏好70% similarVerified从预训练检查点到可商用产品需经历监督式指令微调（SFT）和基于人类反馈的强化学习（RLHF）两个后训练阶段69% similarVerifiedThe three-stage alignment training paradigm of SFT–reward model–RLHF was first systematically proposed by OpenAI in the InstructGPT paper.68% similarUnverifiedThe study starts from the same pretrained GPT-2 Small checkpoint and applies two matched conditions separately: standard continued training and adversarial continued training.68% similarUnverifiedMDT uses LoRA (Low-Rank Adaptation) as its primary fine-tuning method to reduce training cost.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/983001API
curl https://kongchang.com/api/v1/knowledge/claims/983001MCP
get_claim(id=983001)