Abstract:Susceptibilities are a technique for neural network interpretability that studies the response of posterior expectation values of observables to perturbations of the loss. We generalize this construction to the setting of the regret in deep reinforcement learning and investigate the utility of susceptibilities in a simple gridworld model that nevertheless exhibits non-trivial stagewise development. We argue that susceptibilities reveal internal features of the development of the model in parameter space that one cannot detect purely by studying the development of the learned policy. We validate these results with activation-steering, and discuss the framework's extension to RLHF post-training.
| Comments: | 55 pages, comments welcome |
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2605.08007 [cs.LG] |
| (or arXiv:2605.08007v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2605.08007 arXiv-issued DOI via DataCite (pending registration) |
From: Chris Elliott [view email]
[v1]
Fri, 8 May 2026 16:59:38 UTC (19,513 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。