Understanding Helpfulness and Harmless Tension in Reward Models

arXiv:2606.13209v2 Announce Type: replace-cross Abstract: Reward models (RMs) are a key component of reinforcement learning from human feedback, enabling the alignment of large language models with human values such as helpfulness and harmlessness. However, these alignment goals can conflict, leading to interference when optimised jointly. We refer to this conflict as alignment tension and investigate how it manifests in RMs. To this end, we train RMs optimised for helpfulness, harmlessness, and a combined helpful-harmless objective, and compare their behaviours and internal representations. We find that mixed-objective RMs frequently underperform their single-objective counterparts, suggesting substantial interference between alignment goals. Using activation-based interpretability methods, we identify neurons associated with helpfulness and harmlessness and show that they causally promote their corresponding behaviours while often suppressing the opposing one. At the same time, we find that a substantial fraction of neurons are shared across objectives. These shared neurons exert a disproportionate influence on model behaviour and encode incompatible representations. Our findings provide mechanistic evidence for how alignment objectives are represented and interact within RMs, offering new insights into the challenges of multi-objective alignment and motivating the development of more disentangled and controllable alignment methods.

Abstract:Reward models (RMs) are a key component of reinforcement learning from human feedback, enabling the alignment of large language models with human values such as helpfulness and harmlessness. However, these alignment goals can conflict, leading to interference when optimised jointly. We refer to this conflict as alignment tension and investigate how it manifests in RMs. To this end, we train RMs optimised for helpfulness, harmlessness, and a combined helpful-harmless objective, and compare their behaviours and internal representations. We find that mixed-objective RMs frequently underperform their single-objective counterparts, suggesting substantial interference between alignment goals. Using activation-based interpretability methods, we identify neurons associated with helpfulness and harmlessness and show that they causally promote their corresponding behaviours while often suppressing the opposing one. At the same time, we find that a substantial fraction of neurons are shared across objectives. These shared neurons exert a disproportionate influence on model behaviour and encode incompatible representations. Our findings provide mechanistic evidence for how alignment objectives are represented and interact within RMs, offering new insights into the challenges of multi-objective alignment and motivating the development of more disentangled and controllable alignment methods.

Machine Learning (cs.LG); Computation and Language (cs.CL)

arXiv:2606.13209 [cs.LG]

(or arXiv:2606.13209v2 [cs.LG] for this version)

https://doi.org/10.48550/arXiv.2606.13209 Focus to learn more arXiv-issued DOI via DataCite

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

analiza aiqbench

Understanding Helpfulness and Harmless Tension in Reward Models

Subiecte conexe