Testing the Construct Validity of a Functional Valence Axis in LLM Agents

arXiv:2609.22850v1 Announce Type: new Abstract: Contrastive activation directions are often interpreted from what they decode or how strongly they steer behavior. But what evidence is sufficient to identify the construct represented by such a direction, rather than a correlated feature of the contrast used to extract it? We study this question for a good--bad outcome direction in a maze task, using controlled interventions that separate the realised outcome from the informational history through which it became known. Across multiple LLM checkpoints, directions fitted on one explicit outcome encoding transfer well to another, indicating that the readout is not tied to surface form. In contrast, when the same realised outcome is reached through announced and unannounced histories, transfer degrades substantially: even after both histories receive the same explicit outcome, the post-event readout remains strongly conditioned on the earlier announcement. In a matched maze-RL run, the post-RL direction becomes substantially more predictive of reference-MDP remaining return and the policy becomes more dependent on it at the tested sites, while this history dependence persists. These results support a functional, value-related interpretation of the direction, but not its identification with a history-invariant scalar valence state.

Abstract:Contrastive activation directions are often interpreted from what they decode or how strongly they steer behavior. But what evidence is sufficient to identify the construct represented by such a direction, rather than a correlated feature of the contrast used to extract it? We study this question for a good--bad outcome direction in a maze task, using controlled interventions that separate the realised outcome from the informational history through which it became known. Across multiple LLM checkpoints, directions fitted on one explicit outcome encoding transfer well to another, indicating that the readout is not tied to surface form. In contrast, when the same realised outcome is reached through announced and unannounced histories, transfer degrades substantially: even after both histories receive the same explicit outcome, the post-event readout remains strongly conditioned on the earlier announcement. In a matched maze-RL run, the post-RL direction becomes substantially more predictive of reference-MDP remaining return and the policy becomes more dependent on it at the tested sites, while this history dependence persists. These results support a functional, value-related interpretation of the direction, but not its identification with a history-invariant scalar valence state.

Machine Learning (cs.LG); Artificial Intelligence (cs.AI)

arXiv:2609.22850 [cs.LG]

(or arXiv:2609.22850v1 [cs.LG] for this version)

https://doi.org/10.48550/arXiv.2609.22850 Focus to learn more arXiv-issued DOI via DataCite (pending registration)

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

Testing the Construct Validity of a Functional Valence Axis in LLM Agents

Subiecte conexe