Brief

Researchers have developed a new approach called the Generative Response Model (GRM) for auto-bidding systems in advertising. This model predicts future traffic volume and cost/value curves based on historical data and a bid multiplier. Unlike previous methods that integrate constraints into reward signals, GRM directly models responses, which is shown to improve constraint stability and overall performance on the AuctionNet dataset. AI

Researchers have introduced a new method called Reward Modeling from Natural Language Human Feedback (RM-NLHF) to improve the training of Generative Reward Models (GRMs). Traditional methods using pairwise preference data can lead to GRMs learning to guess correct outcomes without genuine understanding, introducing noise into the training signal. RM-NLHF addresses this by using natural language critiques from humans to provide more accurate process reward signals, which are then used to train GRMs. The approach also includes a Meta Reward Model (MetaRM) to generalize from limited human critiques to larger datasets. AI