In-Context Learning Enhanced Credibility Transformer
This work addresses the need for better model learning and predictive performance in AI, particularly for handling new instances with unseen categorical features, though it appears incremental as it builds on an existing Credibility Transformer architecture.
The paper tackles the problem of improving predictive accuracy and generalization in Transformer models by introducing an in-context learning mechanism that augments the Credibility Transformer with a context batch of similar instances, resulting in enhanced CLS token representations and adaptation to similar risk patterns.
The starting point of our network architecture is the Credibility Transformer which extends the classical Transformer architecture by a credibility mechanism to improve model learning and predictive performance. This Credibility Transformer learns credibilitized CLS tokens that serve as learned representations of the original input features. In this paper we present a new paradigm that augments this architecture by an in-context learning mechanism, i.e., we increase the information set by a context batch consisting of similar instances. This allows the model to enhance the CLS token representations of the instances by additional in-context information and fine-tuning. We empirically verify that this in-context learning enhances predictive accuracy by adapting to similar risk patterns. Moreover, this in-context learning also allows the model to generalize to new instances which, e.g., have feature levels in the categorical covariates that have not been present when the model was trained -- for a relevant example, think of a new vehicle model which has just been developed by a car manufacturer.