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

2026-09-19

AI Summary

研究通过控制干预分离结果与信息历史，测试LLM代理中功能效价轴的构建有效性，表明其具有价值相关性但非历史不变。

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.

Problem

• Contrastive activation directions
• construct validity
• functional valence
• LLM agents
• maze task

Innovation

• Contrastive Activation Directions
• Functional Valence
• Outcome Encoding
• Informational History
• Maze Task