Randall Balestriero

Into the Rabbit Hull: From Task-Relevant Concepts in DINO to Minkowski Geometry

OpenApps: Simulating Environment Variations to Measure UI Agent Reliability

LLM-JEPA: Large Language Models Meet Joint Embedding Predictive Architectures

SAFE: Benchmarking AI Weather Prediction Fairness with Stratified Assessments of Forecasts over Earth

Task Priors: Enhancing Model Evaluation by Considering the Entire Space of Downstream Tasks

Next Token Perception Score: Analytical Assessment of your LLM Perception Skills

The Centroid Affinity Hypothesis: How Deep Network Represent Features

What Deep Networks Want to Learn and How to Get There

LoRA Users Beware: A Few Spurious Tokens Can Manipulate Your Finetuned Model

Cross-Entropy Is All You Need To Invert the Data Generating Process

Joint‑Embedding vs Reconstruction: Provable Benefits of Latent Space Prediction for Self‑Supervised Learning

Ditch the Denoiser: Emergence of Noise Robustness in Self-Supervised Learning from Data Curriculum

Curvature Tuning: Provable Training-free Model Steering From a Single Parameter

FastDINOv2: Frequency Based Curriculum Learning Improves Robustness and Training Speed

Learning from Reward-Free Offline Data: A Case for Planning with Latent Dynamics Models

ScaLES: Scalable Latent Exploration Score for Pre-Trained Generative Networks

No Location Left Behind: Measuring and Improving the Fairness of Implicit Representations for Earth Data

$\mathbb{X}$-Sample Contrastive Loss: Improving Contrastive Learning with Sample Similarity Graphs

MITIGATING OVER-EXPLORATION IN LATENT SPACE OPTIMIZATION USING LES

The Fair Language Model Paradox

ALLoRA: Adaptive Learning Rate Mitigates LoRA Fatal Flaws

Eidetic Learning: an Efficient and Provable Solution to Catastrophic Forgetting