Skip to yearly menu bar Skip to main content


Workshop

Actionable Interpretability

Tal Haklay ⋅ Hadas Orgad ⋅ Anja Reusch ⋅ Marius Mosbach ⋅ Sarah Wiegreffe ⋅ Ian Tenney ⋅ Mor Geva
Project Page [ OpenReview] 

Abstract

Video

Chat is not available.

Schedule

Timezone: Asia/Seoul
1:00 AM
Opening Remarks
→ DCBM: Data-Efficient Visual Concept Bottleneck Models
Katharina Prasse ⋅ Patrick Knab ⋅ Sascha Marton ⋅ Christian Bartelt ⋅ Margret Keuper
→ Detecting High-Stakes Interactions with Activation Probes
Alex McKenzie ⋅ Phil Blandfort ⋅ Urja Pawar ⋅ William Bankes ⋅ David Krueger ⋅ Ekdeep Singh Lubana ⋅ Dmitrii Krasheninnikov
→ Simplicity isn't as simple as you think
Aman Sinha ⋅ Timothee Mickus ⋅ Marianne Clausel ⋅ Mathieu Constant ⋅ Xavier Coubez
→ Machine Learning from Explanations
Jiashu Tao ⋅ Reza Shokri
→ MSE-Break: Steering Internal Representations to Bypass Refusals in Large Language Models
Ashwin Saraswatula ⋅ Pranav Balabhadra ⋅ Pranav Dhinakar
→ Challenges in Understanding Modality Conflict in Vision-Language Models
Trang Nguyen ⋅ Jackson Michaels ⋅ Madalina Fiterau ⋅ David Jensen
→ Avoiding Leakage Poisoning: Concept Interventions Under Distribution Shifts
Mateo Espinosa Zarlenga ⋅ Gabriele Dominici ⋅ Pietro Barbiero ⋅ Zohreh Shams ⋅ Mateja Jamnik
→ Probabilistic Soundness Guarantees in LLM Reasoning Chains
Weiqiu You ⋅ Anton Xue ⋅ Shreya Havaldar ⋅ Delip Rao ⋅ Helen Jin ⋅ Chris Callison-Burch ⋅ Eric Wong
→ MIB: A Mechanistic Interpretability Benchmark
Aaron Mueller ⋅ Atticus Geiger ⋅ Sarah Wiegreffe ⋅ Dana Arad ⋅ Iván Arcuschin ⋅ Adam Belfki ⋅ Yik Siu Chan ⋅ Jaden Fiotto-Kaufman ⋅ Tal Haklay ⋅ Michael Hanna ⋅ Jing Huang ⋅ Rohan Gupta ⋅ Yaniv Nikankin ⋅ Hadas Orgad ⋅ Nikhil Prakash ⋅ Anja Reusch ⋅ Aruna Sankaranarayanan ⋅ Shun Shao ⋅ Alessandro Stolfo ⋅ Martin Tutek ⋅ Amir Zur ⋅ David Bau ⋅ Yonatan Belinkov
→ Single Feature Tips the Balance: Reducing Language Model Over-Refusal with Sparse Representations
Ailin Deng ⋅ Shaoliang Nie ⋅ Lijuan Liu ⋅ Xianjun Yang ⋅ Ujjwal Karn ⋅ Dat Huynh ⋅ Fulton Wang ⋅ Ying Xu ⋅ Madian Khabsa ⋅ Saghar Hosseini
→ A Single Direction of Truth: An Observer Model’s Linear Residual Probe Exposes and Steers Contextual Hallucinations
Charles O'Neill ⋅ Sviatoslav Chalnev ⋅ Rune Chi Zhao ⋅ Max Kirkby ⋅ Mudith Jayasekara
→ Interpretable Diffusion Models with B-cos Networks
Nicola Bernold ⋅ Moritz Vandenhirtz ⋅ Alice Bizeul ⋅ Julia Vogt
→ Taming Knowledge Conflicts in Language Models
Gaotang Li ⋅ Yuzhong Chen ⋅ Hanghang Tong
→ Benchmarking Deception Probes via Black-to-White Performance Boosts
Avi Parrack ⋅ Carlo Attubato ⋅ Stefan Heimersheim
→ Steering Language Model Refusal with Sparse Autoencoders
Kyle O'Brien ⋅ David Majercak ⋅ Xavier Fernandes ⋅ Richard Edgar ⋅ Blake Bullwinkel ⋅ Jingya Chen ⋅ Harsha Nori ⋅ Dean Carignan ⋅ Eric Horvitz ⋅ Forough Poursabzi-Sangdeh
→ Classifier Reconstruction Through Counterfactual-Aware Wasserstein Prototypes
Xuan Zhao ⋅ Zhuo Cao ⋅ Arya Bangun ⋅ Hanno Scharr ⋅ Ira Assent
→ Internal Causal Mechanisms Robustly Predict Language Model Out-of-Distribution Behaviors
Jing Huang ⋅ Junyi Tao ⋅ Thomas Icard ⋅ Diyi Yang ⋅ Christopher Potts
→ Learning interpretable positional encodings in transformers depends on initialization
Takuya Ito ⋅ Luca Cocchi ⋅ Tim Klinger ⋅ Parikshit Ram ⋅ Murray Campbell ⋅ Luke Hearne
→ Probing for Arithmetic Errors in Language Models
Yucheng Sun ⋅ Alessandro Stolfo ⋅ Mrinmaya Sachan
→ Insights into a radiology-specialised multimodal large language model with sparse autoencoders
Kenza Bouzid ⋅ Shruthi Bannur ⋅ Felix Meissen ⋅ Daniel Coelho de Castro ⋅ Anton Schwaighofer ⋅ Javier Alvarez-Valle ⋅ Stephanie L Hyland
→ Scalpel vs. Hammer: GRPO Amplifies Existing Capabilities, SFT Replaces Them
Neel Rajani ⋅ Aryo Gema ⋅ Seraphina Goldfarb-Tarrant ⋅ Ivan Titov
→ Looking Beyond the Top-1: Transformers Determine Top Tokens in Order
Daria Lioubashevski ⋅ Tomer Schlank ⋅ Gabriel Stanovsky ⋅ Ariel Goldstein
→ Discovering Forbidden Topics in Language Models
Can Rager ⋅ Chris Wendler ⋅ Rohit Gandikota ⋅ David Bau
→ Beyond Multiple Choice: Evaluating Steering Vectors for Adaptive Free-Form Summarization
Joschka Braun ⋅ Carsten Eickhoff ⋅ Seyed Ali Bahrainian
→ Neural Concept Verifier: Scaling Prover-Verifier Games via Concept Encodings
Berkant Turan ⋅ Suhrab Asadulla ⋅ David Steinmann ⋅ Wolfgang Stammer ⋅ Sebastian Pokutta
→ Steering off Course: Reliability Challenges in Steering Language Models
Patrick Da Silva ⋅ Hari Sethuraman ⋅ Dheeraj Rajagopal ⋅ Hannaneh Hajishirzi ⋅ Sachin Kumar
→ Bayesian Influence Functions for Scalable Data Attribution
Philipp Kreer ⋅ Wilson Wu ⋅ Maxwell Adam ⋅ Zach Furman ⋅ Jesse Hoogland
→ Activation Steering in Generative Settings via Contrastive Causal Mediation Analysis
Aruna Sankaranarayanan ⋅ Amir Zur ⋅ Atticus Geiger ⋅ Dylan Hadfield-Menell
→ Understanding Synthetic Context Extension via Retrieval Heads
Xinyu Zhao ⋅ Fangcong Yin ⋅ Greg Durrett
→ Internal states before wait modulate reasoning patterns
Dmitrii Troitskii ⋅ Koyena Pal ⋅ Chris Wendler ⋅ Callum McDougall ⋅ Neel Nanda
→ Posthoc Disentanglement of Textual and Acoustic Features in Self-Supervised Speech Encoders
Hosein Mohebbi ⋅ Grzegorz Chrupała ⋅ Willem Zuidema ⋅ Afra Alishahi ⋅ Ivan Titov
→ Identifiable Steering via Sparse Autoencoding of Multi-Concept Shifts
Shruti Joshi ⋅ Andrea Dittadi ⋅ Sebastien Lachapelle ⋅ Dhanya Sridhar
→ Interpretable EEG-to-Image Generation with Semantic Prompts
Arshak Rezvani ⋅ Ali Akbari ⋅ Kosar Arani ⋅ Maryam Mirian ⋅ Emad Arasteh ⋅ Martin McKeown
→ Prediction Models That Learn to Avoid Missing Values
Lena Stempfle ⋅ Anton Matsson ⋅ Newton Mwai ⋅ Fredrik Johansson
→ MULTIGUARD: An Efficient Approach for AI Safety Moderation Across Modalities
Sahil Verma ⋅ Keegan Hines ⋅ Jeff Bilmes ⋅ Charlotte Siska ⋅ Luke Zettlemoyer ⋅ Hila Gonen ⋅ Chandan Singh
→ Rethinking Crowd-Sourced Evaluation of Neuron Explanations
Tuomas Oikarinen ⋅ Ge Yan ⋅ Akshay Kulkarni ⋅ Lily Weng
→ Are language models aware of the road not taken? Token-level uncertainty and hidden state dynamics
Amir Zur ⋅ Eric Bigelow ⋅ Atticus Geiger ⋅ Ekdeep Singh Lubana
→ Convergent Linear Representations of Emergent Misalignment
Anna Soligo ⋅ Edward Turner ⋅ Senthooran Rajamanoharan ⋅ Neel Nanda
→ Interpreting the repeated token phenomenon in LLMs
Itay Yona ⋅ Jamie Hayes ⋅ Ilia Shumailov ⋅ Federico Barbero ⋅ Yossi Gandelsman
→ Beyond the ATE: Interpretable Modelling of Treatment Effects over Dose and Time
Julianna Piskorz ⋅ Krzysztof Kacprzyk ⋅ Harry Amad ⋅ Mihaela van der Schaar
→ Diversity-driven Data Selection for Language Model Tuning through Sparse Autoencoder
Xianjun Yang ⋅ Shaoliang Nie ⋅ Lijuan Liu ⋅ Suchin Gururangan ⋅ Ujjwal Karn ⋅ Rui Hou ⋅ Madian Khabsa ⋅ Yuning Mao
→ Towards eliciting latent knowledge from LLMs with mechanistic interpretability
Bartosz Cywiński ⋅ Emil Ryd ⋅ Senthooran Rajamanoharan ⋅ Neel Nanda
→ Explanation Design in Strategic Learning: Sufficient Explanations that Induce Non-harmful Responses
Kiet Vo ⋅ Siu Lun Chau ⋅ Masahiro Kato ⋅ Yixin Wang ⋅ Krikamol Muandet
→ Probing and Steering Evaluation Awareness of Language Models
Jord Nguyen ⋅ Hoang Khiem ⋅ Carlo Attubato ⋅ Felix Hofstätter
→ Sum-of-Parts: Self-Attributing Neural Networks with End-to-End Learning of Feature Groups
Weiqiu You ⋅ Helen Qu ⋅ Marco Gatti ⋅ Bhuvnesh Jain ⋅ Eric Wong
→ Why Do Some Inputs Break Low-Bit LLMs?
Ting-Yun Chang ⋅ Muru Zhang ⋅ Jesse Thomason ⋅ Robin Jia
→ Resilient Multi-Concept Steering in LLMs via Enhanced Sparse "Conditioned" Autoencoders
Saurish Srivastava ⋅ Kevin Zhu ⋅ Cole Blondin ⋅ Sean O'Brien
→ Automated Detection of Visual Attribute Reliance with a Self-Reflective Agent
Christy Li ⋅ Josep Camuñas ⋅ Jake Touchet ⋅ Jacob Andreas ⋅ Agata Lapedriza ⋅ Antonio Torralba ⋅ Tamar Rott Shaham
7:00 AM
Coffee Break
8:00 AM
Panel on Actionable Interpretability
Fazl Barez ⋅ Naomi Saphra ⋅ Samuel Marks ⋅ Kyle Lo
8:45 AM
Closing Remarks
What Kind of User Are You? Uncovering User Models in LLM Chatbots
Yida Chen ⋅ Aoyu Wu ⋅ Trevor DePodesta ⋅ Catherine Yeh ⋅ Lena Armstrong ⋅ Kenneth Li ⋅ Nicholas Marin ⋅ Oam Patel ⋅ Jan Riecke ⋅ Shivam Raval ⋅ Olivia Seow ⋅ Martin Wattenberg ⋅ Fernanda Viégas
Are We Merely Justifying Results ex Post Facto? Quantifying Explanatory Inversion in Post-Hoc Model Explanations
Zhen Tan ⋅ Song Wang ⋅ Yifan Li ⋅ Yu Kong ⋅ Jundong Li ⋅ Tianlong Chen ⋅ huan liu
Learning-Augmented Robust Algorithmic Recourse
Kshitij Kayastha ⋅ Shahin Jabbari ⋅ Vasilis Gkatzelis
ExpProof : Operationalizing Explanations for Confidential Models with ZKPs
Chhavi Yadav ⋅ Evan Laufer ⋅ Dan Boneh ⋅ Kamalika Chaudhuri
Breaking Bad Tokens: Detoxification of LLMs Using Sparse Autoencoders
Agam Goyal ⋅ Vedant Rathi ⋅ William Yeh ⋅ Yian Wang ⋅ Yuen Chen ⋅ Hari Sundaram
Persistent Demographic Information in X-ray Foundation Embeddings: a Risk for a Safe and Fair Deployment in Healthcare
Filipe Santos ⋅ Aldo Marzullo ⋅ Alessandro Quarta ⋅ João Sousa ⋅ Susana Vieira ⋅ Leo Celi ⋅ Francesco Calimeri ⋅ Laleh Seyyed-Kalantari
BlueGlass: A Framework for Composite AI Safety
Harshal Sanjay Nandigramwar ⋅ Qutub Syed Sha ⋅ Kay-Ulrich Scholl
On the Effect of Uncertainty on Layer-wise Inference Dynamics
Sunwoo Kim ⋅ Haneul Yoo ⋅ Alice Oh
MPF: Aligning and Debiasing Language Models post Deployment via Multi-Perspective Fusion
Xin Guan ⋅ Pei-Hsin Lin ⋅ Zekun Wu ⋅ Ze Wang ⋅ Ruibo Zhang ⋅ Emre Kazim ⋅ Adriano Koshiyama
Why Do Metrics Think That? Towards Understanding Large Language Models as Machine Translation Evaluators
Runzhe Zhan ⋅ Xinyi Yang ⋅ Junchao Wu ⋅ Lidia Chao ⋅ Derek Wong
Adversarial Cooperative Rationalization: The Risk of Spurious Correlations in Even Clean Datasets
Wei Liu ⋅ Zhongyu Niu ⋅ Lang Gao ⋅ Zhiying Deng ⋅ Jun Wang ⋅ Haozhao Wang ⋅ Ruixuan Li
Transferring Features Across Language Models With Model Stitching
Alan Chen ⋅ Jack Merullo ⋅ Alessandro Stolfo ⋅ Ellie Pavlick