Matthias Bethge

Full Professor@University of Tuebingen·德国·OpenReview

研究方向

language model agents · open-ended benchmarking · compositional machine learning · lifelong learning · object-centric learning · biological vision · benchmarking · comparing humans to machines · comparing neural representations · brain score · cognitive map learning · modeling the retina

Matthias Bethge

Strategic Dishonesty Can Undermine AI Safety Evaluations of Frontier LLMs

Mapping Post-Training Forgetting in Language Models at Scale

Only Brains Align with Brains: Cross-Region Alignment Patterns Expose Limits of Normative Models

OVid: Open Large-Scale Video Dataset as a Novel Source for Image-Text Data

Un-Attributability: Computing Novelty from Retrieval & Semantic Similarity

Concept-Aware Batch Sampling Improves Language-Image Pretraining

From Slots to Masks: Rethinking OCL

In Search of Forgotten Domain Generalization

WikiBigEdit: Understanding the Limits of Lifelong Knowledge Editing in LLMs

A Sober Look at Progress in Language Model Reasoning: Pitfalls and Paths to Reproducibility

Can Language Models Falsify? Evaluating Algorithmic Reasoning with Counterexample Creation

Great Models Think Alike and this Undermines AI Oversight

Identifying latent state transitions in non-linear dynamical systems

What Moves the Eyes: Doubling Mechanistic Model Performance Using Deep Networks to Discover and Test Cognitive Hypotheses

Equivariance by Contrast: Identifiable Equivariant Embeddings from Unlabeled Finite Group Actions

Testing the Limits of Fine-Tuning for Improving Visual Cognition in Vision Language Models

LLMs on the Line: Data Determines Loss-to-Loss Scaling Laws

Democratizing Evaluation with Infinity-Benchmarks: Sample-Level Heterogeneous Testing Over Arbitrary Capabilities