Implicit vs. Explicit Prompting Strategies for LVLMs in Referential Communication

In three linesTwo recent studies reach contradictory conclusions about LVLMs' ability to coordinate efficient referring expressions. This research controls for task differences and directly compares prompting styles. Models coordinate efficiently with explicit prompting but fail to infer communicative efficiency needs from implicit prompts.

Summary generated by Claude — human-verified