NeurIPS 2025 Past Other
UniReps: 3rd Edition of the Workshop on Unifying Representations in Neural Models
UniReps2025
- Submission deadline
- Aug 30, 2025, 14:20 UTC imported from OpenReview — check the website for extensions
- Submission portal
- OpenReview
- Notes
- Auto-imported from the OpenReview venue record on 2026-06-10 — please verify and enrich (topics are keyword-guessed).
Accepted papers (116)
Fetched from OpenReview (v2) on 2026-06-10.
-
A Circular Argument: Does RoPE need to be Equivariant for Vision?
-
A second-order perspective on linear mode connectivity modulo permutations
-
All In: Bridging Input Feature Spaces Towards Graph Foundation Models
-
An Empirical Study of Task and Feature Correlations in the Reuse of Pre-trained Models
-
An Empirical Study on Unifying JEPA and Language Supervision for Visual Representation Learning
-
Analyze the Neurons, not the Embeddings: Understanding When and Where LLM Representations Align with Humans
-
Any-Subgroup Equivariant Networks via Symmetry Breaking
-
Better Representations, Better Speech BCIs: a Multitask Approach
-
Better Together: Leveraging Unpaired Multimodal Data for Stronger Unimodal Models
-
Beyond [cls]: Exploring the true potential of Masked Image Modeling representations
-
Bias-driven Alignment of Linear and ReLU Networks
-
Blinded by Language: Multimodal LLMs Underuse Their Vision Backbone
-
Brain–Language Model Alignment: Insights into the Platonic Hypothesis and Intermediate-Layer Advantage
-
Bridging Large Gaps in Neural Network Representations with Model Stitching
-
Building expertise through task-specific representational alignment in biological and artificial neural networks
-
Clifford Algebraic Rotor Embeddings : Maybe embeddings should start to CARE
-
Condition-Dependent Representational Alignment between Whisper and the Human Speech Network
-
Context-Aware World Models for Task-Agnostic Control
-
Contrastive Representations for Temporal Reasoning
-
Cross-Modal Representational Alignment with LLM Priors for Image Generation
-
Data Augmentation Techniques to Reverse-Engineer Neural Network Weights from Input-Output Queries
-
Data symmetries generate drifting similarity matrices in manifold-tiling neural codes
-
Dead Feature Counts in Sparse Autoencoders Predict Underlying Deep Q Networks' Effectiveness
-
Decoding Projections From Frozen Random Weights in Autoencoders: What Information Do They Encode?
-
DIET-CP: Lightweight and Data Efficient Self Supervised Continued Pretraining
-
Distinguishing probabilistic from non-probabilistic neural representations
-
Ditch the Denoiser: Emergence of Noise Robustness in Self-Supervised Learning from Data Curriculum
-
DocQIR-Emb: Document Image Retrieval with Multi-lingual Question Query
-
Echo of Bayes: Learned Memory Functions Can Recover Belief States
-
Enhancing Multimodal Product Retrieval in E-Commerce by Reversing Typographic Attacks
-
Equivalences between network modularity and diverse low-dimensional representations
-
Escaping Plato’s Cave: JAM for Aligning Independently Trained Vision and Language Models
-
Evaluating Foundation Models' 3D Understanding Through Multi-View Correspondence Analysis
-
Exact Learning Dynamics of Bottlenecked and Wide Deep Linear Networks
-
Exploring Augmentation-Driven Invariances for Graph Self-supervised Learning in Spatial Omics
-
Finding Fingerprints of Out-Of-Distribution Failures from In-Distribution Geometry
-
From Aggregation to Guidance: Strategies for Personalized Federated Fine-Tuning of Foundation Models
-
GraphMatch: Fusing Language and Graph Representations in a Dynamic Two-Sided Work Marketplace
-
Group Equivariance Meets Mechanistic Interpretability: Equivariant Sparse Autoencoders
-
Hand-Engineered Image-Computable Models Can Still Outperform DNNs in V1 Similarity
-
Human-like individual differences emerge from random weight initializations in neural networks
-
Improving Generation Quality of Long-Tailed Diffusion via Disentangled Latent Representations
-
Interpreting convolutional neural networks to study wide-field amacrine cell inhibition in the retina
-
Learning Resilient Molecular Representations with Dynamic Multi-Modal Fusion
-
Learning using switching synaptic plasticity rules
-
Leveraging Parameter Space Symmetries for Reasoning Skill Transfer in LLMs
-
LevyScore: A Fast Sample-Wise Confidence Score of Pretrained Joint Embedding Model
-
Linear Maps for Cross-Model Finetuning Transfer
-
Linear Recurrent Networks Approximate Optimal Filtering in Hidden Markov Models
-
LLM-JEPA: Large Language Models Meet Joint Embedding Predictive Architectures
-
Look, Then Speak: Social Tokens for Grounding LLMs in Visual Interactions
-
Low-Rank Successor Representations Capture Human-Like Generalization
-
MASS: MoErging through Adaptive Subspace Selection
-
Measure Before You Look: Grounding Embeddings Through Manifold Metrics
-
Measuring and Controlling Solution Degeneracy across Task-Trained Recurrent Neural Networks
-
Measuring the Measures: Discriminative Capacity of Representational Similarity Metrics Across Model Families
-
Message-Passing State-Space Models: Improving Graph Learning with Modern Sequence Modeling
-
Mice to Machines: Neural Representations from Visual Cortex for Domain Generalization
-
Misalignment Between Vision-Language Representations in Vision-Language Models
-
MultiPersona-Align: Zero-Shot Multi-Subject Personalized Image Generation with Layout-Guidance via Dual Representation Alignment
-
Neural Correlates of Language Models Are Specific to Human Language
-
Neural Embedding Alignment Reveals Nonlinear Latent Transformations across Brain Regions
-
NeuroFusion: A Unified Framework for Generalized Visual Stimulus Decoding from fMRI Across Datasets and Subjects
-
Neuron-Level Linguistic Selectivity in LLMs via a Classifier-Free Framework
-
No Clustering, No Routing: How Transformers Actually Process Rare Tokens
-
On Defining Neural Averaging
-
On Task Vectors and Gradients
-
On the Identifiability of Latent Action Policies
-
On the Impact of Topological Regularization on Geometrical and Topological Alignment in Autoencoders: An Empirical Study
-
On the Training Dynamics of Contrastive Learning with Imbalanced Feature Distributions: A Theoretical Study of Feature Learning
-
One Question at a Time: A Semantic Bottleneck for Interpretable Visual Brain Decoding from fMRI
-
Perspective: Summary Statistics of Learning
-
Phase codes emerge in recurrent neural networks optimized for modular arithmetic
-
Quantifying information stored in synaptic connections rather than in firing activities of neural networks
-
Quantum Relational Knowledge Distillation
-
R²-CoD: Understanding Text-Graph Complementarity in Relational Reasoning via Knowledge Co-Distillation
-
Radial-VCReg: More Informative Representation Learning Through Radial Gaussianization
-
Relational Representation Learning
-
Representing Neural Network Layers as Linear Operations via Koopman Operator Theory
-
Rethinking Objectives for Multi-View and Multi-Modal Contrastive Learning
-
Scratchpad Thinking: Alternation Between Storage and Computation in Latent Reasoning Models
-
Self-Supervised Learning from Structural Invariance
-
SemCLIP: A Semantic Memory-Aligned Vision Language Model
-
Shared Parameter Subspaces and Cross-Task Linearity in Emergently Misaligned Behavior
-
Signatures of the Auditory Cortex Reveal Discrepancies Across Speech Recognition Models
-
SoftStep: learning instance-wise similarity functions between neural representations
-
Sparse Autoencoder Neural Operators: Model Recovery in Function Spaces
-
Spectral Insights into Data-Oblivious Critical Layers in Large Language Models
-
Stable Single-Pixel Contrastive Learning for Semantic and Geometric Tasks
-
stable-pretraining: Foundation Model Research Made Simple
-
Superclass-Guided Representation Disentanglement for Spurious Correlation Mitigation
-
Superposition disentanglement of neural representations reveals hidden alignment
-
Superposition in Graph Neural Networks
-
SWAT-NN: Simultaneous Weights and Architecture Training for Neural Networks in a Latent Space
-
Symmetry-Aware Fully-Amortized Optimization with Scale Equivariant Graph Metanetworks
-
Task Priors: Enhancing Model Evaluation by Considering the Entire Space of Downstream Tasks
-
TextIT: Inference-Time Representation Alignment for Improved Visual Text Generation in Diffusion Models
-
The Fragility of Polarity: A Perturbative Analysis of the sign Hypothesis in Sparse Networks
-
The Geometry and Topology of Modular Addition Representations
-
The Performance Cost of Representational Misalignment
-
Topological Alignment of Shared Vision-Language Embedding Space
-
Towards Interpretable Deep Neural Networks for Tabular Data
-
Towards Mitigating Systematics in Large-Scale Surveys via Few-Shot Optimal Transport-Based Feature Alignment
-
Transformers as Unrolled Inference in Probabilistic Laplacian Eigenmaps
-
Two-Scale Latent Dynamics for Recurrent-Depth Transformers
-
Understanding Task Transfer in Vision-Language Models
-
Unifying Vision-Language Latents for Zero-label Image Caption Enhancement
-
Universal Properties of Activation Sparsity in Modern Large Language Models
-
Universally Converging Representations of Matter Across Scientific Foundation Models
-
Unraveling the cognitive patterns of Large Language Models through module communities
-
VISGate: ROI-Conditioned Dual-Head Encoders that Align Visual Features and Brain Responses
-
VQ-Kernels: Unraveling Deep Learning of High-Dimensional Data Geometry
-
Weight Weaving: Parameter Pooling for Data-Free Model Merging
-
Where does an LLM begin computing an instruction?
-
Why and How Auxiliary Tasks Improve JEPA Representations
-
Windsock is Dancing: Adaptive Multimodal Retrieval-Augmented Generation