Production and Perception in LLMs: A Token Probability Approach

2026-07-13

AI Summary

This study investigates whether large language models (LLMs) exhibit a functional asymmetry between generation and perception tasks analogous to human language processing. By designing distinct generative and perceptual prompts, the authors re-score token probabilities for identical input texts and employ prompt reconstruction alongside time-series analysis to systematically compare probability distributions across these two task types in five open-source models, including Llama-3.1-8B. The work demonstrates, for the first time in pure decoder architectures, a stable distinction between generative and perceptual conditions, revealing that the probability distance between them is significantly larger—on average 1.8 times greater—than that between two generative conditions. This effect consistently replicates across models and prompts, providing evidence for task-dependent functional asymmetry within LLMs.

Abstract

The asymmetry between language production and perception has been well-documented in psycholinguistics. Whether large language models (LLMs) exhibit a functionally analogous distinction remains an open question, particularly given that LLMs rely on the same underlying mechanism (next-token prediction) for both input and output processing. In this exploratory study, we operationalize the production-perception distinction through direct token probability measurements rather than metalinguistic prompting. Using the base Llama-3.1-8B model, we generated poems under a production prompt and re-scored the same tokens under both rephrased production prompts and perception-oriented prompts. Across an extended experiment with four production and three perception prompts, production-perception distances consistently and substantially exceeded production-production distances, with non-overlapping ranges across conditions and an overall average ratio of approximately 1.8. Near-ceiling correlations in the production-production control confirm that the effect is specific to communicative framing rather than prompt surface variation, and we show the effect replicates across five open-weight models (Llama-3.1-8B, EuroLLM-9B, gemma-2-9b-it, Mistral-7B-Instruct-v0.3, and Qwen2.5-7B-Instruct), spanning both base and instruction-tuned variants. Temporal analysis revealed that the perception prompt exerts its strongest influence at the beginning of the sequence, with divergence decaying as generated context accumulates, though the specific shape of this decay varies across prompt pairs. These findings suggest that prompt framing alone induces a production-perception distinction in LLM probability distributions, even within a decoder-only architecture.

Problem

• production-perception asymmetry
• large language models
• token probability
• prompt framing
• decoder-only architecture

Innovation

• token probability
• production-perception asymmetry
• prompt framing
• large language models
• decoder-only architecture