Unverified50% confidenceFactExact time
PPLM dynamically adjusts hidden state directions during the generation process by computing the gradient of a classifier with respect to the model's hidden states, steering outputs to better conform to target attributes.
1
Sources
50%
Confidence
Long-term
Relevance
9/2/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedPlug-and-play methods such as PPLM and FUDGE keep the parameters of the pre-trained language model unchanged and, during the decoding stage, adjust the generation probability distribution through additionally trained lightweight classifiers to guide the output to satisfy specific attributes.71% similarUnverified分类模型的推理是非自回归的，仅需一次前向传播即可完成所有位置预测，计算效率显著高于生成模型66% similarVerifiedLIME通过在目标样本周围构造扰动数据集，训练局部可解释的线性模型来近似复杂模型的局部行为63% similarUnverified贝叶斯滤波是POMDP中维护信念状态的核心机制，通过状态转移模型预测先验概率、用观测模型计算似然、归一化得到后验信念63% similarUnverifiedModel steering influences model output behavior during inference by directly modifying the activation vectors of intermediate layers, without modifying weights.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/845569API
curl https://kongchang.com/api/v1/knowledge/claims/845569MCP
get_claim(id=845569)