# Metonymic Circuits for Abstract Concept Grounding in Vision Transformers

arXiv: https://arxiv.org/abs/2610.06928 · HTML: https://mlsanity.com/paper/2610.06928/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 25 of its 72 references that are in the map.
- First posted on arXiv 3 Oct 2026.

## Builds on (most cited first, 25 of 25)

- [Locating and Editing Factual Associations in GPT](https://mlsanity.com/paper/2202.05262/) (2022)
- [Sparse Autoencoders Find Highly Interpretable Features in Language Models](https://mlsanity.com/paper/2309.08600/) (2023)
- [Shikra: Unleashing Multimodal LLM's Referential Dialogue Magic](https://mlsanity.com/paper/2306.15195/) (2023)
- [Progress measures for grokking via mechanistic interpretability](https://mlsanity.com/paper/2301.05217/) (2023)
- [Ferret: Refer and Ground Anything Anywhere at Any Granularity](https://mlsanity.com/paper/2310.07704/) (2023)
- [GLaMM: Pixel Grounding Large Multimodal Model](https://mlsanity.com/paper/2311.03356/) (2023)
- [The Neuro-Symbolic Concept Learner: Interpreting Scenes, Words, and Sentences From Natural Supervision](https://mlsanity.com/paper/1904.12584/) (2019)
- [Transcoders Find Interpretable LLM Feature Circuits](https://mlsanity.com/paper/2406.11944/) (2024)
- [Interpreting CLIP's Image Representation via Text-Based Decomposition](https://mlsanity.com/paper/2310.05916/) (2023)
- [Grounded Situation Recognition](https://mlsanity.com/paper/2003.12058/) (2020)
- [Interpreting CLIP with Hierarchical Sparse Autoencoders](https://mlsanity.com/paper/2502.20578/) (2025)
- [Steering CLIP's vision transformer with sparse autoencoders](https://mlsanity.com/paper/2504.08729/) (2025)
- [Exploring Visual Interpretability for Contrastive Language-Image Pre-training](https://mlsanity.com/paper/2209.07046/) (2022)
- [Decomposing and Interpreting Image Representations via Text in ViTs Beyond CLIP](https://mlsanity.com/paper/2406.01583/) (2024)
- [DECEPTICON: How Dark Patterns Manipulate Web Agents](https://mlsanity.com/paper/2512.22894/) (2025)
- [It's a (Blind) Match! Towards Vision-Language Correspondence without Parallel Data](https://mlsanity.com/paper/2503.24129/) (2025)
- [Beyond the Doors of Perception: Vision Transformers Represent Relations Between Objects](https://mlsanity.com/paper/2406.15955/) (2024)
- [Grammar-Based Grounded Lexicon Learning](https://mlsanity.com/paper/2202.08806/) (2022)
- [Probing the Representational Power of Sparse Autoencoders in Vision Models](https://mlsanity.com/paper/2508.11277/) (2025)
- [Circuit Tracing in Vision-Language Models: Understanding the Internal Mechanisms of Multimodal Thinking](https://mlsanity.com/paper/2602.20330/) (2026)
- [VisualTrap: A Stealthy Backdoor Attack on GUI Agents via Visual Grounding Manipulation](https://mlsanity.com/paper/2507.06899/) (2025)
- [The Mechanistic Emergence of Symbol Grounding in Language Models](https://mlsanity.com/paper/2510.13796/) (2025)
- [Causal Interpretation of Sparse Autoencoder Features in Vision](https://mlsanity.com/paper/2509.00749/) (2025)
- [Context-based Word Acquisition for Situated Dialogue in a Virtual World](https://mlsanity.com/paper/1401.6875/) (2014)
- [A Computational Approach to Visual Metonymy](https://mlsanity.com/paper/2601.17706/) (2026)

## Abstract (arXiv)

We study how Vision Transformers ground abstract concepts (e.g., angry) when training data provide limited direct referential evidence. We hypothesize a metonymic grounding mechanism in which abstract predictions are driven by concrete, interpretable anchor concepts (e.g., fire) that bridge visual signals to abstract semantics. By applying Transcoders on CLIP and DINO vision encoders, we recover intermediate features that can be associated with semantic labels for more concrete concepts, and trace their contributions in circuits underlying abstract concept recognition. Experiments on a carefully curated icon dataset reveal structured metonymic circuits, in which perceptual primitives dominate early layers and object-like anchors precede abstract targets. Images containing rendered text instead recruit a distinct perceptual-to-textual route. Causal interventions further validate that metonymic intermediates are functionally involved in grounding abstract concepts.