DB
David Bau
cs.LGcs.CVcs.CLcs.AIcs.GRcs.CYstat.MLcs.NEcs.SEcomputer vision
On Valency
published · living versionsW_wjvdyh8g·v1 · currentpublished
Elucidating the Design Space of LM Data Attribution
with Steven Yu, Sheng-Yu Wang, Kimia Hamidieh, Antonio Torralba
1 version
Preprints & journals
81 papers in the corpus · 2017–2026The Persona Hierarchy Model: Understanding Contextual Generalization in Fine-Tuning LLMs2610.09384v1 · Jiachen Zhao, Zhengxuan Wu, David Bau et al.2026 · 0 citationsarXiv
How to train your model organism2610.10203v1 · Xilin Wang, David Bau, Byron C. Wallace2026 · 0 citationsarXiv
Identifying Introspection From the Inside2610.07186v1 · David I. Atkinson, Dillon Plunkett, David Bau2026 · 0 citationsarXiv
Learning the identity: a case study of how SGD selects among functional decompositions2610.00615v1 · Andy Arditi, Weian Xie, David Bau et al.2026 · 0 citationsarXiv
Using OCR Heads to Verbalize Image Semantics2609.18823v1 · Sheridan Feucht, Benno Krojer, Sarah Wang et al.2026 · 0 citationsarXiv
Mechanistic Interpretability of Structure-Aware Numerical Reasoning in LLaMA 3.1 8B2608.18419v1 · Rahul Chowdhury, Timothy A Rupprecht, Senhao Cao et al.2026 · 0 citationsarXiv
Gaze Heads: How VLMs Look at What They Describe2606.14703v1 · Rohit Gandikota, David Bau2026 · 0 citationsarXiv
The Piggyback Hypothesis of Generalization: Explaining and Mitigating Emergent Misalignment2606.06667v3 · Jiachen Zhao, Zhengxuan Wu, Aryaman Arora et al.2026 · 0 citationsarXiv
Shared Lexical Task Representations Explain Behavioral Variability In LLMs2604.22027v2 · Zhuonan Yang, Jacob Xiaochen Li, Francisco Piedrahita Velez et al.2026 · 0 citationsarXiv
The Dual Mechanisms of Spatial Variable Binding in Vision-Language Models2603.22278v3 · Kelly Cui, Nikhil Prakash, Shoval Messica et al.2026 · 0 citationsarXiv
Agents of Chaos2602.20021v1 · Natalie Shapira, Chris Wendler, Avery Yen et al.2026 · 7 citationsarXiv
Two Stages of Folding: Convergent Mechanisms in AI Protein Folding Trunks2602.06020v3 · Kevin Lu, Jannik Brinkmann, Stefan Huber et al.2026 · 1 citationarXiv
Do explanations generalize across large reasoning models?2601.11517v1 · Koyena Pal, David Bau, Chandan Singh2026 · 0 citationsarXiv
In-Context Algebra2512.16902v2 · Eric Todd, Jannik Brinkmann, Rohit Gandikota et al.2025 · 0 citationsarXiv
Vector Arithmetic in Concept and Token Subspaces2511.18162v1 · Sheridan Feucht, Byron Wallace, David Bau2025 · 0 citationsarXiv
LLMs Process Lists With General Filter Heads2510.26784v3 · Arnab Sen Sharma, Giordano Rogers, Natalie Shapira et al.2025 · 0 citationsarXiv
LLMs Encode Harmfulness and Refusal Separately2507.11878v5 · Jiachen Zhao, Jing Huang, Zhengxuan Wu et al.2025 · 1 citationarXiv
Discovering Forbidden Topics in Language Models2505.17441v3 · Can Rager, Chris Wendler, Rohit Gandikota et al.2025 · 0 citationsarXiv
When Are Concepts Erased From Diffusion Models?2505.17013v5 · Kevin Lu, Nicky Kriplani, Rohit Gandikota et al.2025 · 4 citationsAdvances in Neural Information Processing Systems 38 (NeurIPS 2025), 2025
Language Models use Lookbacks to Track Beliefs2505.14685v3 · Nikhil Prakash, Natalie Shapira, Arnab Sen Sharma et al.2025 · 0 citationsarXiv
Leveraging AI for Productive and Trustworthy HPC Software: Challenges and Research Directions2505.08135v1 · Keita Teranishi, Harshitha Menon, William F. Godoy et al.2025 · 4 citationsIn: Neuwirth, S., Paul, A.K., Weinzierl, T., Carson, E.C. (eds) High Performance Computing. ISC High Performance 2025. Lecture Notes in Computer Science, vol 16091. Springer, Cham
MIB: A Mechanistic Interpretability Benchmark2504.13151v2 · Aaron Mueller, Atticus Geiger, Sarah Wiegreffe et al.2025 · 1 citationarXiv
The Dual-Route Model of Induction2504.03022v2 · Sheridan Feucht, Eric Todd, Byron Wallace et al.2025 · 0 citationsarXiv
Distilling Diversity and Control in Diffusion Models2503.10637v4 · Rohit Gandikota, David Bau2025 · 4 citationsarXiv
Elucidating Mechanisms of Demographic Bias in LLMs for Healthcare2502.13319v2 · Hiba Ahsan, Arnab Sen Sharma, Silvio Amir et al.2025 · 1 citationarXiv
Position-aware Automatic Circuit Discovery2502.04577v1 · Tal Haklay, Hadas Orgad, David Bau et al.2025 · 2 citationsarXiv
SliderSpace: Decomposing the Visual Capabilities of Diffusion Models2502.01639v1 · Rohit Gandikota, Zongze Wu, Richard Zhang et al.2025 · 6 citationsarXiv
Open Problems in Mechanistic Interpretability2501.16496v1 · Lee Sharkey, Bilal Chughtai, Joshua Batson et al.2025 · 11 citationsarXiv
Machine Unlearning Doesn't Do What You Think: Lessons for Generative AI Policy and Research2412.06966v2 · A. Feder Cooper, Christopher A. Choquette-Choo, Miranda Bogen et al.2024 · 3 citationsarXiv
Opt-In Art: Learning Art Styles Only from Few Examples2412.00176v4 · Hui Ren, Joanna Materzynska, Rohit Gandikota et al.2024 · 0 citationsarXiv
One-Step is Enough: Sparse Autoencoders for Text-to-Image Diffusion Models2410.22366v5 · Viacheslav Surkov, Chris Wendler, Antonio Mari et al.2024 · 1 citationarXiv
Erasing Conceptual Knowledge from Language Models2410.02760v3 · Rohit Gandikota, Sheridan Feucht, Samuel Marks et al.2024 · 1 citationarXiv
The Quest for the Right Mediator: Surveying Mechanistic Interpretability Through the Lens of Causal Mediation Analysis2408.01416v3 · Aaron Mueller, Jannik Brinkmann, Millicent Li et al.2024 · 2 citationsarXiv
Measuring Progress in Dictionary Learning for Language Model Interpretability with Board Game Models2408.00113v2 · Adam Karvonen, Benjamin Wright, Can Rager et al.2024 · 1 citationarXiv
Open Problems in Technical AI Governance2407.14981v2 · Anka Reuel, Ben Bucknall, Stephen Casper et al.2024 · 9 citationsTransactions on Machine Learning Research, 2025
NNsight and NDIF: Democratizing Access to Open-Weight Foundation Model Internals2407.14561v4 · Jaden Fiotto-Kaufman, Alexander R. Loftus, Eric Todd et al.2024 · 4 citationsarXiv
Token Erasure as a Footprint of Implicit Vocabulary Items in LLMs2406.20086v3 · Sheridan Feucht, David Atkinson, Byron Wallace et al.2024 · 5 citationsarXiv
Customizing Text-to-Image Models with a Single Image Pair2405.01536v2 · Maxwell Jones, Sheng-Yu Wang, Nupur Kumari et al.2024 · 24 citationsarXiv
Locating and Editing Factual Associations in Mamba2404.03646v2 · Arnab Sen Sharma, David Atkinson, David Bau2024 · 3 citationsarXiv
Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models2403.19647v3 · Samuel Marks, Can Rager, Eric J. Michaud et al.2024 · 4 citationsInternational Conference on Learning Representations, 2025
Model Lakes2403.02327v2 · Koyena Pal, David Bau, Ren'ee J. Miller2024 · 0 citationsarXiv
Fine-Tuning Enhances Existing Mechanisms: A Case Study on Entity Tracking2402.14811v1 · Nikhil Prakash, Tamar Rott Shaham, Tal Haklay et al.2024 · 3 citationsarXiv
Measuring and Controlling Instruction (In)Stability in Language Model Dialogs2402.10962v4 · Kenneth Li, Tianle Liu, Naomi Bashkansky et al.2024 · 4 citationsarXiv
Black-Box Access is Insufficient for Rigorous AI Audits2401.14446v3 · Stephen Casper, Carson Ezell, Charlotte Siegmann et al.2024 · 79 citationsThe 2024 ACM Conference on Fairness, Accountability, and Transparency (FAccT '24), June 3-6, 2024, Rio de Janeiro, Brazil
Concept Sliders: LoRA Adaptors for Precise Control in Diffusion Models2311.12092v2 · Rohit Gandikota, Joanna Materzynska, Tingrui Zhou et al.2023 · 41 citationsarXiv
An Alternative to Regulation: The Case for Public AI2311.11350v1 · Nicholas Vincent, David Bau, Sarah Schwettmann et al.2023 · 0 citationsarXiv
Testing Language Model Agents Safely in the Wild2311.10538v3 · Silen Naihin, David Atkinson, Marc Green et al.2023 · 4 citationsarXiv
Future Lens: Anticipating Subsequent Tokens from a Single Hidden State2311.04897v1 · Koyena Pal, Jiuding Sun, Andrew Yuan et al.2023 · 9 citationsarXiv
Function Vectors in Large Language Models2310.15213v2 · Eric Todd, Millicent L. Li, Arnab Sen Sharma et al.2023 · 5 citationsarXiv
FIND: A Function Description Benchmark for Evaluating Interpretability Methods2309.03886v3 · Sarah Schwettmann, Tamar Rott Shaham, Joanna Materzynska et al.2023 · 2 citationsNeurIPS 2023
Career total: 145 works. 81 are in this corpus.Showing the 50 most recent.
Profile built from the corpus for this byline.
Author records are still filling in while the Hub is in alpha. If this is your page, you'll be able to claim it soon. Spot a mistake? Tell us.