mlsanityData to 14 Oct 2026

Vision-Language Models are Fragile Multilingual Associators

arXiv 2608.12333 ↗First posted 2 Jun 2026cs.CL cs.AI cs.CVcited by 0

On the map: Multimodal LLMs / vision-language models → VLM benchmarks and visual reasoning

Abstract (arXiv)

Vision-language models must associate visual entities with textual attributes. Whether these associations or concept bindings remain stable when the language of the input changes is unexplored. We introduce M2BIND, a benchmark varying the language of the context and query across multiple languages. We evaluate binding both extrinsically through task performance metrics and intrinsically through causal interventions. We find that binding is not language-invariant: cross-family and cross-script settings trigger significant binding collapse, with the model's internal binding computation shifting to later layers and losing causal strength. Closely related languages preserve associations comparatively better. In a broader sense, our findings indicate how VLMs deployed globally in multilingual settings cannot be assumed to maintain the same association quality observed in monolingual evaluation.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in VLM benchmarks and visual reasoning, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 15 of its 41 references that are in the map. First posted on arXiv 2 Jun 2026.

Builds on

15 of 41 references.

Closest ancestors · 2

Same direction, newest first

  1. 2025Investigating Mechanisms for In-Context Vision Language Binding28 May
  2. 2024CVQA: Culturally-diverse Multilingual Visual Question Answering Benchmark10 Jun

Other references · 13

  1. 2019Unsupervised Cross-lingual Representation Learning at Scalecited by 3,935
  2. 2022Locating and Editing Factual Associations in GPTcited by 2,691
  3. 2019How multilingual is Multilingual BERT?cited by 645
  4. 2020How Good is Your Tokenizer? On the Monolingual Performance of Multilingual Language Modelscited by 318
  5. 2025LLaVA-OneVision-1.5: Fully Open Framework for Democratized Multimodal Trainingcited by 244
  6. 2023Language Model Tokenizers Introduce Unfairness Between Languagescited by 214
  7. 2023Linearity of Relation Decoding in Transformer Language Modelscited by 200
  8. 2024Fine-Tuning Enhances Existing Mechanisms: A Case Study on Entity Trackingcited by 149
  9. 2023How do Language Models Bind Entities in Context?cited by 108
  10. 2024Pangea: A Fully Open Multilingual Multimodal LLM for 39 Languagescited by 64
  11. 2023Discovering Variable Binding Circuitry with Desideratacited by 34
  12. 2024Representational Analysis of Binding in Language Modelscited by 28
1 more
  1. 2025Mixing Mechanisms: How Language Models Retrieve Bound Entities In-Contextcited by 26

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2024Understanding the Limits of Vision Language Models Through the Lens of the Binding Problem
  2. 2026Seeing to Generalize: How Visual Data Corrects Binding Shortcuts
  3. 2026The Dual Mechanisms of Spatial Variable Binding in Vision-Language Models
  4. 2025V-SEAM: Visual Semantic Editing and Attention Modulating for Causal Interpretability of Vision-Language Models
  5. 2026The Geometry of Representational Failures in Vision Language Models
  6. 2025Can VLMs Recall Factual Associations From Visual References?
  7. 2025Words or Vision: Do Vision-Language Models Have Blind Faith in Text?
  8. 2026Diagnosing the Sources of Compositional Failure in Vision-Language Models: A Controlled Analysis