It Takes Two: Complementary Self-Distillation for Contextual Integrity in LLMs

CRISP: Compositional Reasoning over Images via Stackable Programs for VLMs

MEME: Multi-Entity & Evolving Memory Evaluation

Half-Truths Break Similarity-Based Retrieval

LLM generation novelty through the lens of semantic similarity

How can embedding models bind concepts?

When Do Diffusion Models learn to Generate Multiple Objects?

Compositional Generalization Requires Linear, Orthogonal Representations in Vision Embedding Models

Universal Algorithm-Implicit Learning

Dynamics Reveals Structure: Challenging the Linear Propagation Assumption

MASEval: Extending Multi-Agent Evaluation from Models to Systems

Privacy Collapse: Benign Fine-Tuning Can Break Contextual Privacy in Language Models

CLIP Behaves like a Bag-of-Words Model Cross-modally but not Uni-modally

DISCO: Diversifying Sample Condensation for Efficient Model Evaluation

Dr.LLM: Dynamic Layer Routing for LLMs

Enhancing Multi-Image Understanding through Delimiter Token Scaling

SelfReflect: Can LLMs Communicate Their Internal Answer Distribution?

Break the Output Geometry for Large Language Model Unlearning

Lost in Communication: Uncertainty Propagation in Multi-Agent Systems

CLIP Models Generalize Less Than Compositional Benchmarks Suggest

Agent in a Bottle: Can LLM Agents Turn Their Capabilities Into Cheap, Scalable Artifacts?

Form Over Content In Gradient-Based Data Attribution Methods

EgoGapBench: Benchmarking Egocentric Action Selection in Multi-Agent Scenes

MuPPET: A Benchmark for Contextual Privacy of LLM Assistants in Multi-Party Conversations

Predicting Future Behaviors in Reasoning Models Enables Better Steering

Sparse Autoencoders enable Robust and Interpretable Fine-tuning of CLIP models

Diffusion Classifiers Understand Compositionality, but Conditions Apply

On the Rankability of Visual Embeddings

OVS Meets Continual Learning: Towards Sustainable Open-Vocabulary Segmentation

C-SEO Bench: Does Conversational SEO Work?

Leaky Thoughts: Large Reasoning Models Are Not Private Thinkers

Does Data Scaling Lead to Visual Compositional Generalization?

Do Deep Neural Network Solutions Form a Star Domain?

Intermediate Layer Classifiers for OOD Generalization

Decoupled Finetuning for Domain Generalizable Semantic Segmentation

Playing repeated games with Large Language Models

Are We Done with Object-Centric Learning?

Mitigating Shortcut Learning with Diffusion Counterfactuals and Diverse Ensembles

DiCoTTA: Domain-invariant Learning for Continual Test-time Adaptation

Benchmarking Uncertainty Disentanglement: Specialized Uncertainties for Specialized Tasks

Scaling Up Membership Inference: When and How Attacks Succeed on Large Language Models

Studying Large Language Model Behaviors Under Realistic Knowledge Conflicts

Calibrating Large Language Models Using Their Generations Only

TRAP: Targeted Random Adversarial Prompt Honeypot for Black-Box Identification

Towards User-Focused Research in Training Data Attribution for Human-Centered Explainable AI

Scalable Ensemble Diversification for OOD Generalization and Detection

Pretrained Visual Uncertainties

A Bayesian Perspective On Training Data Attribution

ID and OOD Performance Are Sometimes Inversely Correlated on Real-world Datasets

URL: A Representation Learning Benchmark for Transferable Uncertainty Estimates

ProPILE: Probing Privacy Leakage in Large Language Models

Neglected Free Lunch -- Learning Image Classifiers Using Annotation Byproducts

Scratching Visual Transformer's Back with Uniform Attention

Probabilistic Contrastive Learning Recovers the Correct Aleatoric Uncertainty of Ambiguous Inputs

Exploring Practitioner Perspectives On Training Data Attribution Explanations

URL: A Representation Learning Benchmark for Transferable Uncertainty Estimates

Trustworthy Machine Learning

ECCV Caption: Correcting False Negatives by Collecting Machine-and-Human-verified Image-Caption Associations for MS-COCO

Dataset Condensation via Efficient Synthetic-Data Parameterization

Weakly Supervised Semantic Segmentation Using Out-of-Distribution Data

Which Shortcut Cues Will DNNs Choose? A Study from the Parameter-Space Perspective