Preference
RLHF · PBRL | B-Pref:生成多样非理性 preference,建立 PBRL benchmark
贡献:提出一种生成非理性(模拟人类)preference 的方法,使用多样化的 preference,评测了 PBRL 各环节算法设计(select informative queries、feedback schedule)的效果。 ......
RLHF · PBRL | PEBBLE:通过 human preference 学习 reward model
① 使用熵 intrinsic reward 的 agent pre-training,② 选择尽可能 informative 的 queries 去获取 preference,③ 使用更新后的 reward model 对 replay buffer 进行 relabel。 ......
Learning Heterogeneous Temporal Patterns of User Preference for Timely Recommendation
目录概符号说明TimelyRecMulti-aspect Time Encoder (MATE)Time-aware History Encoder (TAHE)Prediction代码 Cho J., Hyun D., Kang S. and Yu H. Learning heterogeneou ......
Measuring the diversity of recommendations: a preference-aware approach for evaluating and adjusting diversity
Meymandpour R. and Davis J. G. Measuring the diversity of recommendations: a preference-aware approach for evaluating and adjusting diversity. Knowled ......