Jun 2026· arXiv.org· Vol abs/2606.17372· 0 citations· 53 references
Computer Science
Abstract
Two recent studies \citep{jones2026llms, zeng2026lvlms} reach apparently contradictory conclusions about whether large vision-language models (LVLMs) can coordinate similarly to humans 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 \textit{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 \textit{implicit} prompt, highlighting critical differences between how humans and AI systems communicate.
Extensive evaluations show that stereotype-conditioned prompting substantially outperforms generic baselines across all three languages, obtaining significant gains in factuality, specificity, cogency, and effectiveness for both explicit and implicit implied stereotypes.
Greta Damo, Elias Urios Alacreu, Elena Cabrio et al.· 0 citations
Adults generate causally-structured situation models during comprehension and use them to predict upcoming referents. In an eye-tracked comprehension task, 36-month-olds (N = 115) did the same, using story-provided information to predict which character could plausibly be mentioned next. In critical sentences such as “...
We present a formal analysis of an understudied postverbal completive marker -le in Mandarin, distinct from the perfective -le. We propose that completive -le forces strict telicity via maximal affectedness of the theme. Building on degree-based approaches, we show that -le interacts with a covert partitive operator to...
Yenan Sun, Lawrence Y. L. Cheung, Jackie Yan-Ki Lai· Natural language & linguisti...· 0 citations
Learning language requires learning not only the content of language, but also how to use language to communicate. Iterated reference games provide a window into such skills, requiring sustained interactive communication as participants converge on mutually understandable names for initially novel referents. Some ear...
Veronica Boyce, Robert Z. Sparks, Michael C. Frank· Open Mind· 0 citations
The complexity of ensuring consistent behavior as models become increasingly multimodal and retrieval-augmented is illustrated, showing that the longer processing time associated with resolving visual entities prevents the suppression of the model's usual factual recall mechanism, thus resulting in more parametric answ...
Taking a new look at some widely accepted analyses of such syntactic patterns as minimality, head-movement, and LF, this book offers alternative theories while still working within the general Universal Grammar framework. It is the first of its kind to present an explicit and candid examination of how Motivated Reasoni...
