Unverified50% confidenceBenchmarkExact time
斯坦福CRFM团队的HELM基准测试发现，经过RLHF对齐的模型在创意写作任务上的词汇多样性指标普遍低于对应的基础预训练模型
1
Sources
50%
Confidence
Long-term
Relevance
7/15/2026
First Seen
Sources
口语化采样登顶ICML：提示工程配得上顶会吗？
Related Claims
Unverified大语言模型的指令跟随能力是通过RLHF训练习得的统计倾向，注意力机制无法从架构层面区分硬规则与软建议75% similarUnverified提示词与模型版本存在深度绑定关系，不同版本模型在RLHF训练数据、指令微调策略上的差异会导致相同提示词产生不同输出，这种现象被称为「提示词漂移」75% similarUnverified模型在未收到明确指令时输出图片，意味着系统提示词设计或RLHF训练中存在激励偏差75% similarUnverifiedResearchers have attempted to improve models' pragmatic understanding through instruction tuning and reinforcement learning from human feedback (RLHF), but in essence this amounts to layering heuristic rules on top of statistical models.74% similarUnverifiedA 2023 paper by OpenAI researchers found that models trained with RLHF tend to express agreement even when faced with clearly incorrect views (sycophancy bias).72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/510730API
curl https://kongchang.com/api/v1/knowledge/claims/510730MCP
get_claim(id=510730)