Search papers, labs, and topics across Lattice.
This paper introduces ENTRAP-VL, a taxonomically structured instrument designed to investigate contextual entrainment in vision-language models (VLMs). By recognizing that contextual entrainment in VLMs is a dual phenomenon influenced by both textual and visual contexts, the authors provide a curated dataset of 1,500 items across eight categories to facilitate rigorous exploration of this behavior. The findings underscore the necessity of a dedicated framework for studying contextual influences in multimodal settings, which has implications for the reliability and interpretability of VLM outputs.
Contextual entrainment in vision-language models reveals a duality that could fundamentally alter our understanding of multimodal interactions.
Contextual entrainment is the tendency of a model to let auxiliary context in its input pull its output, independently of whether that context is relevant, true, or even meaningful. Recently, it has been identified and given a mechanistic account in unimodal language models. Whether and how it manifests in vision-language models (VLMs) is, by contrast, largely unexamined, and the field lacks a purpose-built instrument with which to investigate it. We take the position that studying contextual entrainment in VLMs requires more than porting an existing text-only benchmark to the multimodal setting: it requires a taxonomically structured, dual-modality instrument whose conditions are constructed around the item at hand (the depicted image in the textual stream, the textual query in the visual stream). We argue that the move to VLMs is substantive rather than incremental. It makes entrainment a dual phenomenon, drivable independently by textual and by visual context, and it opens a veracity distinction (context that is false of the depicted scene yet possible in the world) that has no counterpart in the unimodal, world-knowledge-only formulation of prior work. To make this position concrete and actionable, we introduce ENTRAP-VL (ENTRainment Assessment Probe for Vision and Language), a manually curated dataset of 1,500 items across eight categories, organized by a taxonomy that spans two axes, i.e., the association of context with the item and its relationship to truth, and split into a textual-entrainment stream (eight context conditions) and a visual-entrainment stream (three context conditions). We do not claim to measure entrainment in any particular model; we provide the instrument, the taxonomy that motivates it, and the evaluation protocols it enables, so that the community can investigate the phenomenon rigorously. We will release the dataset and its documentation publicly.
