Circuit Tracing in Vision–Language Models: Understanding the Internal Mechanisms of Multimodal Thinking
This is experimental HTML to improve accessibility. We invite you to report rendering errors. Use Alt+Y to toggle on accessible reporting links and Alt+Shift+Y to toggle off. Learn more about this project and help improve conversions. Vision–language models (VLMs) are powerful but remain opaque black boxes. We introduce the first framework for transparent circuit tracing in VLMs to systematically analyze multimodal reasoning. By utilizing transcoders, attribution graphs, and attention-based methods, we uncover how VLMs hierarchically integrate visual and semantic concepts. We reveal that distinct visual feature circuits can handle mathematical reasoning and support cross-modal associations. Validated through feature steering and circuit patching, our framework proves these circuits are causal and controllable, laying the groundwork for more explainable and reliable VLMs. Our code and models are available at https://github.com/UIUC-MONET/vlm-circuit-tracing. The rapid advancement of vi
Tianhu Xiong Email: mw34@illinois.edu Shengyi Qian Email: klara@illinois.edu Klara Nahrstedt Affiliation: University of Illinois Urbana-Champaign Mingyuan Wu Affiliation: University of Illinois Urbana-Champaign Affiliation: Independent Researcher Abstract Vision–language models (VLMs) are powerful but remain opaque black boxes. We introduce the first framework for transparent circuit tracing in VLMs to systematically analyze multimodal reasoning. By utilizing transcoders, attribution graphs, and attention-based methods, we uncover how VLMs hierarchically integrate visual and…
saved by
related reading
- Circuit Tracing: Revealing Computational Graphs in Language Modelstransformer-circuits.pub
- Language Model Circuits Are Sparse in the Neuron Basis | Transluce AItransluce.org
- On the Biology of a Large Language Modeltransformer-circuits.pub
- A Mathematical Framework for Transformer Circuitstransformer-circuits.pub
- Verbalizable Representations Form a Global Workspace in Language Modelstransformer-circuits.pub
- Transformer Circuits Threadtransformer-circuits.pub
- Towards Monosemanticity: Decomposing Language Models With Dictionary Learningtransformer-circuits.pub
- Tracing the thoughts of a large language model \ Anthropicanthropic.com
- Towards Automated Circuit Discovery for Mechanistic Interpretabilityarxiv.org
- [2406.11944] Transcoders Find Interpretable LLM Feature Circuitsarxiv.org
- Towards Monosemanticity: Decomposing Language Models With Dictionary Learningtransformer-circuits.pub
- Bridging the Attention Gap: Complete Replacement Models for Complete Circuit Tracinginterp.open-moss.com