Workshop
UniReps: Unifying Representations in Neural Models
Marco Fumero 路 Zorah L盲hner 路 Luca Moschella 路 Cl茅mentine Domin茅 路 Donato Crisostomi 路 Kimberly Stachenfeld
West Exhibition Hall C, B3
Sat 14 Dec, 8:15 a.m. PST
Neural models tend to learn similar representations when subject to similar stimuli; this behavior has been observed both in biological and artificial settings. The emergence of these similar representations is igniting a growing interest in the fields of neuroscience and artificial intelligence. To gain a theoretical understanding of this phenomenon, promising directions include: analyzing the learning dynamics and studying the problem of identifiability in the functional and parameter space. This has strong consequences in unlocking a plethora of applications in ML from model fusion, model stitching, to model reuse and in improving the understanding of biological and artificial neural models, including large retrained foundation models. The objective of the workshop is to discuss theoretical findings, empirical evidence and practical applications of this phenomenon, benefiting from the cross-pollination of different fields (ML, Neuroscience, Cognitive Science) to foster the exchange of ideas and encourage collaborations. Overall the questions we aim to investigate are when, why and how internal representations of distinct neural models can be unified into a common representation.
Schedule
Sat 8:15 a.m. - 8:30 a.m.
|
Opening Remarks
(
Presentation
)
>
SlidesLive Video |
馃敆 |
Sat 8:30 a.m. - 9:00 a.m.
|
Learning dynamics describe the computational role of (universal) representations
(
Invited Talk
)
>
SlidesLive Video |
Erin Grant 馃敆 |
Sat 9:00 a.m. - 9:30 a.m.
|
Interpretable Theories for Comparing Biological and Artificial Neural Networks
(
Invited Talk
)
>
SlidesLive Video |
SueYeon Chung 馃敆 |
Sat 9:30 a.m. - 10:00 a.m.
|
The Platonic Representation Hypothesis
(
Invited Talk
)
>
SlidesLive Video |
Phillip Isola 馃敆 |
Sat 10:00 a.m. - 10:30 a.m.
|
Coffee Break
|
馃敆 |
Sat 10:30 a.m. - 11:45 a.m.
|
Contributed talks
SlidesLive Video |
馃敆 |
Sat 11:45 a.m. - 12:45 p.m.
|
Panel Discussion
SlidesLive Video |
Alex Williams 路 Matthew Leavitt 路 Erin Grant 馃敆 |
Sat 12:45 p.m. - 2:00 p.m.
|
Lunch
|
馃敆 |
Sat 2:00 p.m. - 2:30 p.m.
|
On Using Optimal Transport Distances to Regularize Representation Learning
(
Invited Talk
)
>
SlidesLive Video |
Marco Cuturi 馃敆 |
Sat 2:30 p.m. - 3:00 p.m.
|
Open Problems in Universality: A mechanistic interpretability perspective
(
Invited Talk
)
>
SlidesLive Video |
Neel Nanda 馃敆 |
Sat 3:00 p.m. - 3:30 p.m.
|
Parameter Symmetries in Neural Networks: what they affect and what to do with them
(
Invited Talk
)
>
SlidesLive Video |
Stefanie Jegelka 馃敆 |
Sat 3:30 p.m. - 3:45 p.m.
|
Closing Remarks
(
Talk
)
>
SlidesLive Video |
馃敆 |
Sat 3:45 p.m. - 5:00 p.m.
|
Poster session
|
馃敆 |
Sat 5:00 p.m. - 5:30 p.m.
|
Social Event
|
馃敆 |
-
|
Equivalence between representational similarity analysis, centered kernel alignment, and canonical correlations analysis ( Poster ) > link | Alex Williams 馃敆 |
-
|
From Lazy to Rich: Exact Learning Dynamics in Deep Linear Networks ( Poster ) > link | Cl茅mentine Domin茅 路 Nicolas Anguita 路 Alexandra Proca 路 Lukas Braun 路 Daniel Kunin 路 Pedro A.M Mediano 路 Andrew Saxe 馃敆 |
-
|
Decision-margin consistency: a principled metric for human and machine performance alignment ( Poster ) > link | George Alvarez 路 Talia Konkle 馃敆 |
-
|
Winning Tickets from Random Initialization: Aligning Masks for Sparse Training ( Poster ) > link | Rohan Jain 路 Mohammed Adnan 路 Ekansh Sharma 路 Yani Ioannou 馃敆 |
-
|
A Framework for Standardizing Similarity Measures in a Rapidly Evolving Field ( Poster ) > link | Nathan Cloos 路 Guangyu Robert Yang 路 Christopher Cueva 馃敆 |
-
|
Position: Maximizing Neural Regression Scores May Not Identify Good Models of the Brain ( Poster ) > link | Rylan Schaeffer 路 Mikail Khona 路 Sarthak Chandra 路 Mitchell Ostrow 路 Brando Miranda 路 Sanmi Koyejo 馃敆 |
-
|
M3CoL: Harnessing Shared Relations via Multimodal Mixup Contrastive Learning for Multimodal Classification ( Poster ) > link | Raja Kumar 路 Raghav Singhal 路 Pranamya Kulkarni 路 Deval Mehta 路 Kshitij Jadhav 馃敆 |
-
|
Understanding Variational Autoencoders with Intrinsic Dimension and Information Imbalance ( Poster ) > link | Charles Camboulin 路 Diego Doimo 路 Aldo Glielmo 馃敆 |
-
|
Investigating the role of modality and training objective on representational alignment between transformers and the brain ( Poster ) > link | Willow Han 路 Ruchira Dhar 路 Qingqing Yang 路 Maryam Behbahani 路 Maria Alejandra Martinez Ortiz 路 Tolulope Oladele 路 Diana C Dima 路 Hsin-Hung Li 路 Anders S酶gaard 路 Yalda Mohsenzadeh 馃敆 |
-
|
Federated GNNs for EEG-Based Stroke Assessment ( Poster ) > link | Andrea Protani 路 Lorenzo Giusti 路 Albert Sund Aillet 路 Simona Sacco 路 Paolo Manganotti 路 Lucio Marinelli 路 Diogo Reis Santos 路 Pierpaolo Brutti 路 Pietro Caliandro 路 Luigi Serio 馃敆 |
-
|
Language decoding from human brain activity via contrastive learning ( Poster ) > link | Matteo Ferrante 路 Nicola Toschi 路 Alexander Huth 馃敆 |
-
|
An Information Criterion for Controlled Disentanglement of Multimodal Data ( Poster ) > link | Chenyu Wang 路 Sharut Gupta 路 Xinyi Zhang 路 Sana Tonekaboni 路 Stefanie Jegelka 路 Tommi Jaakkola 路 Caroline Uhler 馃敆 |
-
|
Comparing the local information geometry of image representations ( Poster ) > link | David Lipshutz 路 Jenelle Feather 路 Sarah Harvey 路 Alex Williams 路 Eero Simoncelli 馃敆 |
-
|
Symmetry Discovery in Neural Network Parameter Spaces ( Poster ) > link | Bo Zhao 路 Nima Dehmamy 路 Robin Walters 路 Rose Yu 馃敆 |
-
|
Representation with a capital 'R' ( Poster ) > link | Jacob Prince 路 George Alvarez 路 Talia Konkle 馃敆 |
-
|
Vision and language representations in multimodal AI models and human social brain regions during natural movie viewing ( Poster ) > link | Hannah Small 路 Haemy Lee Masson 路 Leyla Isik 馃敆 |
-
|
A Cognitive Framework for Learning Debiased and Interpretable Representations via Debiasing Global Workspace ( Poster ) > link | Jinyung Hong 路 Eun Som Jeon 路 Changhoon Kim 路 Keun Hee Park 路 Utkarsh Nath 路 'YZ' Yezhou Yang 路 Pavan Turaga 路 Theodore P. Pavlic 馃敆 |
-
|
Unsupervised Learning of Categorical Structure ( Poster ) > link | Matteo Alleman 路 Stefano Fusi 馃敆 |
-
|
Topology Preserving Regularization for Independent Training of Inter-operable Models ( Poster ) > link | Nicolas Zilberstein 路 Akshay Malhotra 路 Shahab Hamidi-Rad 路 Yugeswar Deenoo 馃敆 |
-
|
Random Propagations in GNNs ( Poster ) > link | Thu Bui 路 Anugunj Naman 路 Carola-Bibiane Sch枚nlieb 路 Bruno Ribeiro 路 Beatrice Bevilacqua 路 Moshe Eliasof 馃敆 |
-
|
Modern Hopfield Networks meet Encoded Neural Representations - Addressing Practical Considerations ( Poster ) > link | Satyananda Kashyap 路 Niharika DSouza 路 Luyao Shi 路 Ken C. L. Wong 路 Hongzhi Wang 路 Tanveer Syeda-Mahmood 馃敆 |
-
|
Hypernetworks for image recontextualization ( Poster ) > link | Maciej Zieba 路 Jakub Balicki 路 Tomasz Dr贸偶d偶 路 Konrad Karanowski 路 Pawel Lorek 路 Hong Lyu 路 Aleksander Skorupa 路 Tomasz Trzcinski 路 Oriol Caudevilla 路 Jakub Tomczak 馃敆 |
-
|
What Representational Similarity Measures Imply about Decodable Information ( Poster ) > link | Sarah Harvey 路 David Lipshutz 路 Alex Williams 馃敆 |
-
|
Connecting Neural Models Latent Geometries with Relative Geodesic Representations ( Poster ) > link | Hanlin Yu 路 Berfin Inal 路 Marco Fumero 馃敆 |
-
|
CopRA: A Progressive LoRA Training Strategy ( Poster ) > link | Zhan Zhuang 路 XIEQUN WANG 路 Yulong Zhang 路 Wei Li 路 Yu Zhang 路 Ying Wei 馃敆 |
-
|
DFM: Interpolant-free Dual Flow Matching ( Poster ) > link | Denis Gudovskiy 路 Tomoyuki Okuno 路 Yohei Nakata 馃敆 |
-
|
Look-Ahead Selective Plasticity for Continual Learning of Visual Tasks ( Poster ) > link | Rouzbeh Meshkinnejad 路 Jie Mei 路 Zeduo Zhang 路 Daniel Lizotte 路 Yalda Mohsenzadeh 馃敆 |
-
|
Emergence of Text Semantics in CLIP Image Encoders ( Poster ) > link | Sreeram Vennam 路 Shashwat Singh 路 Anirudh Govil 路 Ponnurangam Kumaraguru 馃敆 |
-
|
Hybrid Dynamic High-Order Functional Correlations and Divisive Normalization for Improved Classification of Schizophrenia and Bipolar Disorder ( Poster ) > link | Qiang Li 路 Vince Calhoun 馃敆 |
-
|
Video decoding from human fMRI data with a multi-stream sensory approach ( Poster ) > link | Matteo Ferrante 路 Matteo Ciferri 路 Nicola Toschi 馃敆 |
-
|
Invariant Learning with Annotation-free Environments ( Poster ) > link | Phuong Quynh Le 路 J枚rg Schl枚tterer 路 Christin Seifert 馃敆 |
-
|
Workshop Submission: Towards Making Untrainable Networks Trainable ( Poster ) > link | Vighnesh Subramaniam 路 Tomaso Poggio 路 Boris Katz 路 Brian Cheung 路 Andrei Barbu 馃敆 |
-
|
Circular Learning Provides Biological Plausibility ( Poster ) > link | Amin Tavakoli 路 ian domingo 路 Pierre Baldi 馃敆 |
-
|
Auxiliary objectives improve generalization performance but reduce model specification for low-data neuroimaging-based brain age prediction ( Poster ) > link | Donghyun Kim 路 Eloy Geenjaar 路 Vince Calhoun 馃敆 |
-
|
Evidence from fMRI Supports a Two-Phase Abstraction Process in Language Models ( Poster ) > link | Richard Antonello 路 Emily Cheng 馃敆 |
-
|
Understanding Memorization using Representation Similarity Analysis and Model Stitching ( Poster ) > link | Aishwarya Gupta 路 Indranil Saha 路 Piyush Rai 馃敆 |
-
|
Correlating Variational Autoencoders Natively For Multi-View Imputation ( Poster ) > link | Ella Orme 路 Marina Evangelou 路 Ulrich Paquet 馃敆 |
-
|
On the cognitive alignment between humans and machines ( Poster ) > link | Marco Rothermel 路 Soroush Daftarian 路 Tahmineh A. Koosha 路 Mohammad-Ali Nikouei Mahani 路 Hamidreza Jamalabadi 馃敆 |
-
|
Joint Learning for Visual Reconstruction from the Brain Activity: Hierarchical Representation of Image Perception with EEG-Vision Transformer ( Poster ) > link | Ali Ackbari 路 Kosar Arani 路 Tony Muhammad Yousefnezhad 路 Maryam Mirian 路 Emad Arasteh 馃敆 |
-
|
Challenges in Explaining Representational Similarity through Identifiability ( Poster ) > link | Beatrix M. G. Nielsen 路 Luigi Gresele 路 Andrea Dittadi 馃敆 |
-
|
Multi-task Learning yields Disentangled World Models: Impact and Implications ( Poster ) > link | Pantelis Vafidis 路 Aman Bhargava 路 Antonio Rangel 馃敆 |
-
|
Improving OOD Generalization of Pre-trained Encoders via Aligned Embedding-Space Ensembles ( Poster ) > link | Shuman Peng 路 Arash Khoeini 路 Sharan Vaswani 路 Martin Ester 馃敆 |
-
|
Relative Representations: Topological and Geometric Perspectives ( Poster ) > link | Alejandro Garc铆a-Castellanos 路 Giovanni Luca Marchetti 路 Danica Kragic 路 Martina Scolamiero 馃敆 |
-
|
Fast Imagic: Solving Overfitting in Text-guided Image Editing via Disentangled UNet with Forgetting Mechanism and Unified Vision-Language Optimization ( Poster ) > link | Shiwen Zhang 馃敆 |
-
|
Disentangling the Effects of Data Augmentation and Format Transform in Self-Supervised Learning of Image Representations ( Poster ) > link | Neha Kalibhat 路 Warren Morningstar 路 Alex Bijamov 路 Luyang Liu 路 Karan Singhal 路 Philip Mansfield 馃敆 |
-
|
It's All Relative: Relative Uncertainty in Latent Spaces using Relative Representations ( Poster ) > link | Fabian Mager 路 Valentino Maiorca 路 Lars Kai Hansen 馃敆 |
-
|
Inducing Human-like Biases in Moral Reasoning Language Models ( Poster ) > link | Austin Meek 路 Artem Karpov 路 Seong Cho 路 Raymond Koopmanschap 路 Lucy Farnik 路 Bogdan-Ionut Cirstea 馃敆 |
-
|
Revealing spatial-frequency channels in an ensemble encoding model of human fMRI ( Poster ) > link | Furkan Ozcelik 路 Ajay Subramanian 路 Najib Majaj 路 Denis Pelli 馃敆 |
-
|
VISTA: A Panoramic View of Neural Representations ( Poster ) > link | Tom White 馃敆 |
-
|
Representation Learning of Structured Data for Medical Foundation Models ( Poster ) > link | Vijay Prakash Dwivedi 路 Viktor Schlegel 路 Andy Liu 路 Thanh-Tung Nguyen 路 Abhinav Ramesh Kashyap 路 Jeng Wei 路 Wei-Hsian Yin 路 Stefan Winkler 路 Robby Tan 馃敆 |
-
|
Unifying Causal Representation Learning with the Invariance Principle ( Poster ) > link | Dingling Yao 路 Dario Rancati 路 Riccardo Cadei 路 Marco Fumero 路 Francesco Locatello 馃敆 |
-
|
Unsupervised Modality Adaptation in Human Action Recognition via Cross-modal Representation Learning ( Poster ) > link | Abhi Kamboj 路 Duy Nguyen 路 Minh Do 馃敆 |
-
|
Understanding Task Knowledge Entanglement in Protein Language Model Representations ( Poster ) > link | Ria Vinod 馃敆 |
-
|
Monkey See, Model Knew: Large Language Models accurately Predict Human AND Macaque Visual Brain Activity ( Poster ) > link | Colin Conwell 路 Emalie McMahon 路 Akshay Jagadeesh 路 Kasper Vinken 路 Saloni Sharma 路 Jacob Prince 路 George Alvarez 路 Talia Konkle 路 Leyla Isik 路 Margaret Livingstone 馃敆 |
-
|
Rethinking Fine-tuning Through Geometric Perspective ( Poster ) > link | Krishna Sri Ipsit Mantri 路 Moshe Eliasof 路 Carola-Bibiane Sch枚nlieb 路 Bruno Ribeiro 馃敆 |
-
|
Improving Model Merging with Natural Niches ( Poster ) > link | Jo茫o Abrantes 路 Robert Lange 路 Yujin Tang 馃敆 |
-
|
Adapter to facilitate Foundation Model Communication for DLO Instance Segmentation ( Poster ) > link | Omkar Joglekar 路 Shir Simon (Kozlovsky) 路 Dotan Di Castro 馃敆 |
-
|
Self-Supervised Pre-training of Spiking Neural Networks by Contrasting Events and Frames ( Poster ) > link | Raghav Singhal 路 Jan Finkbeiner 路 Emre Neftci 馃敆 |
-
|
Understanding Permutation Based Model Merging with Feature Visualizations ( Poster ) > link | Congshu Zou 路 Geraldin Nanfack 路 Stefan Horoi 路 Eugene Belilovsky 馃敆 |
-
|
DIETing: Self-Supervised Learning with Instance Discrimination Learns Identifiable Features ( Poster ) > link | Attila Juhos 路 Alice Bizeul 路 Patrik Reizinger 路 David Klindt 路 Randall Balestriero 路 Mark Ibrahim 路 Julia Vogt 路 Wieland Brendel 馃敆 |
-
|
Comparing Representations in Static and Dynamic Vision Models to the Human Brain ( Poster ) > link | Hamed Karimi 路 Stefano Anzellotti 馃敆 |
-
|
Locality-aware Concept Bottleneck Model ( Poster ) > link | sujin jeon 路 Inwoo Hwang 路 Sanghack Lee 路 Byoung-Tak Zhang 馃敆 |
-
|
Artificial Neural Networks Explain Continuous Speech Perception in Humans ( Poster ) > link | Gasser Elbanna 路 Josh McDermott 馃敆 |
-
|
Multimodal Lego: Model Merging and Fine-Tuning Across Topologies and Modalities ( Poster ) > link | Konstantin Hemker 路 Nikola Simidjievski 路 Mateja Jamnik 馃敆 |
-
|
Conic Activation Functions ( Poster ) > link | Changqing Fu 路 Laurent Cohen 馃敆 |
-
|
Shared Recurrent Memory Improves Multi-agent Pathfinding ( Poster ) > link | Alsu Sagirova 路 Yury Kuratov 路 Mikhail Burtsev 馃敆 |
-
|
Task-Relevant Covariance from Manifold Capacity Theory Improves Robustness in Deep Networks ( Poster ) > link | William Yang 路 Chi-Ning Chou 路 SueYeon Chung 馃敆 |
-
|
Small-scale adversarial perturbations expose differences between predictive encoding models of human fMRI responses ( Poster ) > link | Nikolas McNeal 路 Mainak Deb 路 N Apurva Ratan Murty 馃敆 |
-
|
Delays in generalization match delayed changes in representational geometry ( Poster ) > link | Xingyu Zheng 路 Kyle Daruwalla 路 Ari Benjamin 路 David Klindt 馃敆 |
-
|
Equivalence between representational similarity analysis, centered kernel alignment, and canonical correlations analysis ( Oral ) > link | Alex Williams 馃敆 |
-
|
A Framework for Standardizing Similarity Measures in a Rapidly Evolving Field ( Oral ) > link | Nathan Cloos 路 Guangyu Robert Yang 路 Christopher Cueva 馃敆 |
-
|
An Information Criterion for Controlled Disentanglement of Multimodal Data ( Oral ) > link | Chenyu Wang 路 Sharut Gupta 路 Xinyi Zhang 路 Sana Tonekaboni 路 Stefanie Jegelka 路 Tommi Jaakkola 路 Caroline Uhler 馃敆 |
-
|
Representation with a capital 'R' ( Oral ) > link | Jacob Prince 路 George Alvarez 路 Talia Konkle 馃敆 |
-
|
Modern Hopfield Networks meet Encoded Neural Representations - Addressing Practical Considerations ( Oral ) > link | Satyananda Kashyap 路 Niharika DSouza 路 Luyao Shi 路 Ken C. L. Wong 路 Hongzhi Wang 路 Tanveer Syeda-Mahmood 馃敆 |
-
|
What Representational Similarity Measures Imply about Decodable Information ( Oral ) > link | Sarah Harvey 路 David Lipshutz 路 Alex Williams 馃敆 |
-
|
Evidence from fMRI Supports a Two-Phase Abstraction Process in Language Models ( Oral ) > link | Richard Antonello 路 Emily Cheng 馃敆 |
-
|
Multimodal Lego: Model Merging and Fine-Tuning Across Topologies and Modalities ( Oral ) > link | Konstantin Hemker 路 Nikola Simidjievski 路 Mateja Jamnik 馃敆 |
-
|
Invited Talk: M.Cuturi
(
Talk
)
>
|
馃敆 |