Instructions Shape Production of Language, not Processing

TL;DR

This paper demonstrates that instructions primarily influence the production stage in language models, with internal information in output tokens correlating strongly with behavior, unlike in sample tokens.

Contribution

It reveals an asymmetry between processing and production stages in language models, emphasizing the importance of analyzing output tokens for understanding capabilities.

Findings

Instruction flow affects output tokens and behavior significantly.

Model scale and instruction tuning sharpen the production-stage effects.

The asymmetry between input processing and output production generalizes across models.

Abstract

Instructions trigger a production-centered mechanism in language models. Through a cognitively inspired lens that separates language processing and production, we reveal this mechanism as an asymmetry between the two stages by probing task-specific information layer-wise across five binary judgment tasks. Specifically, we measure how instruction tokens shape information both when sample tokens, the input under evaluation, are processed and when output tokens are produced. Across prompting variations, task-specific information in sample tokens remains largely stable and correlates only weakly with behavior, whereas the same information in output tokens varies substantially and correlates strongly with behavior. Attention-based interventions confirm this pattern causally: blocking instruction flow to all subsequent tokens reduces both behavior and information in output tokens, whereas blocking it only to sample tokens has minimal effect on either. The asymmetry generalizes across model families and tasks, and becomes sharper with model scale and instruction-tuning, both of which disproportionately affect the production stage. Our findings suggest that understanding model capabilities requires jointly assessing internals and behavior, while decomposing the internal perspective by token position to distinguish the processing of input tokens from the production of output tokens.