David Bau
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
David Bau has a total of 89 co-authors. Here are the top co-authors by number of shared publications:
- Jun-Yan Zhu
Carnegie Mellon University
5 shared publications - Sheng-Yu Wang
Carnegie Mellon University, Pittsburgh, United States of America
5 shared publications Alex Andonian
MIT CSAIL
2 shared publications- Antonio Torralba
CSAIL, MIT, Cambridge, Massachusetts, USA
2 shared publications - Bolei Zhou
University of California, Los Angeles, CA, USA
2 shared publications - Imke Grabe
IT University of Copenhagen, Copenhagen, Denmark
2 shared publications
Latest Publications
- Open Access
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-9
Mechanistic Interpretability of Structure-Aware Numerical Reasoning in LLaMA 3.1–8B
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-9
Open AccessMechanistic Interpretability of Structure-Aware Numerical Reasoning in LLaMA 3.1–8B
- Open Access
- 13 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 594-607
Finding the Unicorn within a Million Patches: A Material-Driven Journey into the Hidden Layers of a Diffusion Model
Today's AI models learn rich internal representations, such as the visual features inside diffusion models that produce generated images, and offer a new kind of material for co-creation. However, interfaces for creating with generative AI typically operate on the level of inputs and outputs, obscuring the material formation that unfolds in between. To address this gap, this pictorial documents a material-driven journey of designing an interface for interacting with the hidden layers of diffusion models. Working in an interdisciplinary team of interpretability researchers and interaction designers, we explore how methods from mechanistic interpretability can inform alternative strategies that move beyond black-boxed interaction paradigms. Through visual exploration of technical material characteristics combining experimentation and drawing, we articulate AI's internal representations as design material. Rather than hiding model properties behind conversational agents, we present a material-driven journey for building interfaces that exposes and engages with them. Finally, we discuss how this approach might enable different forms of co-creation with generative models.
- 13 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 594-607
Open AccessFinding the Unicorn within a Million Patches: A Material-Driven Journey into the Hidden Layers of a Diffusion Model
Today's AI models learn rich internal representations, such as the visual features inside diffusion models that produce generated images, and offer a new kind of material for co-creation. However, interfaces for creating with generative AI typically operate on the level of inputs and outputs, obscuring the material formation that unfolds in between. To address this gap, this pictorial documents a material-driven journey of designing an interface for interacting with the hidden layers of diffusion models. Working in an interdisciplinary team of interpretability researchers and interaction designers, we explore how methods from mechanistic interpretability can inform alternative strategies that move beyond black-boxed interaction paradigms. Through visual exploration of technical material characteristics combining experimentation and drawing, we articulate AI's internal representations as design material. Rather than hiding model properties behind conversational agents, we present a material-driven journey for building interfaces that exposes and engages with them. Finally, we discuss how this approach might enable different forms of co-creation with generative models.
- 24 November 2025
- Published by Springer Nature
- p. 615-625
Leveraging AI for Productive and Trustworthy HPC Software: Challenges and Research Directions
- 24 November 2025
- Published by Springer Nature
- p. 615-625
Leveraging AI for Productive and Trustworthy HPC Software: Challenges and Research Directions
- Open Access
- 18 August 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-5
Hidden Layers: An Interactive Installation for Exploring the Neural Semantics of Image Synthesis
- 18 August 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-5
Open AccessHidden Layers: An Interactive Installation for Exploring the Neural Semantics of Image Synthesis
- Open Access
- 3 December 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-13
Customizing Text-to-Image Models with a Single Image Pair
- 3 December 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-13
Open AccessCustomizing Text-to-Image Models with a Single Image Pair
- Top Cited Papers
- 10 November 2024
- Published by Springer Nature
- p. 172-188
Concept Sliders: LoRA Adaptors for Precise Control in Diffusion Models
- 10 November 2024
- Published by Springer Nature
- p. 172-188
Top Cited PapersConcept Sliders: LoRA Adaptors for Precise Control in Diffusion Models
- Top Cited PapersOpen Access
- 3 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 2254-2272
Black-Box Access is Insufficient for Rigorous AI Audits
External audits of AI systems are increasingly recognized as a key mechanism for AI governance. The effectiveness of an audit, however, depends on the degree of access granted to auditors. Recent audits of state-of-the-art AI systems have primarily relied on black-box access, in which auditors can only query the system and observe its outputs. However, white-box access to the system’s inner workings (e.g., weights, activations, gradients) allows an auditor to perform stronger attacks, more thoroughly interpret models, and conduct fine-tuning. Meanwhile, outside-the-box access to training and deployment information (e.g., methodology, code, documentation, data, deployment details, findings from internal evaluations) allows auditors to scrutinize the development process and design more targeted evaluations. In this paper, we examine the limitations of black-box audits and the advantages of white- and outside-the-box audits. We also discuss technical, physical, and legal safeguards for performing these audits with minimal security risks. Given that different forms of access can lead to very different levels of evaluation, we conclude that (1) transparency regarding the access and methods used by auditors is necessary to properly interpret audit results, and (2) white- and outside-the-box access allow for substantially more scrutiny than black-box access alone.
- 3 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 2254-2272
Top Cited PapersOpen AccessBlack-Box Access is Insufficient for Rigorous AI Audits
External audits of AI systems are increasingly recognized as a key mechanism for AI governance. The effectiveness of an audit, however, depends on the degree of access granted to auditors. Recent audits of state-of-the-art AI systems have primarily relied on black-box access, in which auditors can only query the system and observe its outputs. However, white-box access to the system’s inner workings (e.g., weights, activations, gradients) allows an auditor to perform stronger attacks, more thoroughly interpret models, and conduct fine-tuning. Meanwhile, outside-the-box access to training and deployment information (e.g., methodology, code, documentation, data, deployment details, findings from internal evaluations) allows auditors to scrutinize the development process and design more targeted evaluations. In this paper, we examine the limitations of black-box audits and the advantages of white- and outside-the-box audits. We also discuss technical, physical, and legal safeguards for performing these audits with minimal security risks. Given that different forms of access can lead to very different levels of evaluation, we conclude that (1) transparency regarding the access and methods used by auditors is necessary to properly interpret audit results, and (2) white- and outside-the-box access allow for substantially more scrutiny than black-box access alone.
- Open Access
- 10 December 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-12
Content-based Search for Deep Generative Models
The growing proliferation of customized and pretrained generative models has made it infeasible for a user to be fully cognizant of every model in existence. To address this need, we introduce the task of content-based model search: given a query and a large set of generative models, finding the models that best match the query. As each generative model produces a distribution of images, we formulate the search task as an optimization problem to select the model with the highest probability of generating similar content as the query. We introduce a formulation to approximate this probability given the query from different modalities, e.g., image, sketch, and text. Furthermore, we propose a contrastive learning framework for model retrieval, which learns to adapt features for various query modalities. We demonstrate that our method outperforms several baselines on Generative Model Zoo, a new benchmark we create for the model retrieval task.
- 10 December 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-12
Open AccessContent-based Search for Deep Generative Models
The growing proliferation of customized and pretrained generative models has made it infeasible for a user to be fully cognizant of every model in existence. To address this need, we introduce the task of content-based model search: given a query and a large set of generative models, finding the models that best match the query. As each generative model produces a distribution of images, we formulate the search task as an optimization problem to select the model with the highest probability of generating similar content as the query. We introduce a formulation to approximate this probability given the query from different modalities, e.g., image, sketch, and text. Furthermore, we propose a contrastive learning framework for model retrieval, which learns to adapt features for various query modalities. We demonstrate that our method outperforms several baselines on Generative Model Zoo, a new benchmark we create for the model retrieval task.
- Open Access
An Alternative to Regulation: The Case for Public AI
Can governments build AI? In this paper, we describe an ongoing effort to develop ``public AI'' -- publicly accessible AI models funded, provisioned, and governed by governments or other public bodies. Public AI presents both an alternative and a complement to standard regulatory approaches to AI, but it also suggests new technical and policy challenges. We present a roadmap for how the ML research community can help shape this initiative and support its implementation, and how public AI can complement other responsible AI initiatives.
Open AccessAn Alternative to Regulation: The Case for Public AI
Can governments build AI? In this paper, we describe an ongoing effort to develop ``public AI'' -- publicly accessible AI models funded, provisioned, and governed by governments or other public bodies. Public AI presents both an alternative and a complement to standard regulatory approaches to AI, but it also suggests new technical and policy challenges. We present a roadmap for how the ML research community can help shape this initiative and support its implementation, and how public AI can complement other responsible AI initiatives.
- Open Access
Mass-Editing Memory in a Transformer
Recent work has shown exciting promise in updating large language models with new memories, so as to replace obsolete information or add specialized knowledge. However, this line of work is predominantly limited to updating single associations. We develop MEMIT, a method for directly updating a language model with many memories, demonstrating experimentally that it can scale up to thousands of associations for GPT-J (6B) and GPT-NeoX (20B), exceeding prior work by orders of magnitude. Our code and data are at https://memit.baulab.info.
Open AccessMass-Editing Memory in a Transformer
Recent work has shown exciting promise in updating large language models with new memories, so as to replace obsolete information or add specialized knowledge. However, this line of work is predominantly limited to updating single associations. We develop MEMIT, a method for directly updating a language model with many memories, demonstrating experimentally that it can scale up to thousands of associations for GPT-J (6B) and GPT-NeoX (20B), exceeding prior work by orders of magnitude. Our code and data are at https://memit.baulab.info.