2 papers
cs.LG2026
Training Language Models to Fail Gracefully
Daman Arora, Xuecheng Liu, Himanshu Gaurav Singh +6
Reinforcement learning makes language models better at solving tasks, but the same mechanism also rewards guessing. When success is the only way to earn reward, opting for a safe f…
physics.ao-ph2026
Unmasking Compensating Errors in Numerical Weather Models: A Process-Additive Neural Network Approach
Yuiko Ichikawa, Edward Groot, Jack Atkinson
This work introduces a Process Additive Neural Network (PANN) framework to disentangle compensating biases in atmospheric models, where errors in different physical processes can o…