Quantitatively, we show a variety of bounds relating the number of mistakes to well-known combinatorial dimensions.

Forsuch models, one particularly elegant approach is that ofpassive-aggressive learning[3]. In this framework, a model isonly updated when itfails to classify an example correctly with high confidence.

We consider the problem of a mediator learning to coordinate a group of strategic agents via recommendations of actions to take without knowledge of their underlying utility functions (e.g.

However, the inferred reward function often fails to capture the underlying task objective.

Unfortunately, existing faithfulness criteria for PRP are limited to coarse-grained LLM-based scoring without a clear definition or formulation.

Our code is available at: https://github.com/LINs-lab/ReLA .