Implicit vs. Explicit Prompting Strategies for LVLMs in Referential Communication

2026-06-15

AI Summary

This study investigates whether large vision-language models (LVLMs) can autonomously infer communicative efficiency demands from context and generate efficient referring expressions, as humans do. By systematically comparing explicit and implicit prompting strategies within a unified task framework, we reproduce and integrate the paradigms of Jones et al. and Zeng et al., conducting controlled experiments across multiple LVLMs. The results demonstrate that LVLMs produce efficient referring expressions only under explicit instructions, failing entirely in implicit conditions, thereby revealing a fundamental lack of spontaneous pragmatic reasoning. This work resolves contradictory conclusions in prior literature and provides the first systematic quantification of the pragmatic gap between current AI systems and human performance in referring expression generation.

Abstract

Two recent studies (Jones et al. (2026); Zeng et al. (2026)) reach apparently contradictory conclusions about whether LVLMs can coordinate on efficient referring expressions. We control for task differences between the studies while directly comparing their prompting styles. We replicate the finding that models can coordinate efficient referring expressions when explicitly prompted to do so, suggesting that other task differences are not responsible for divergent results. However, we also find that the same models fail to infer the need for communicative efficiency from a more implicit prompt, highlighting critical differences between how humans and AI systems communicate.

Problem

• referential communication
• large vision-language models
• prompting strategies
• communicative efficiency
• implicit prompting

Innovation

• referential communication
• prompting strategies
• LVLMs
• communicative efficiency
• implicit vs. explicit prompting