7 papers · 1 filter
Towards Understanding Subliminal Learning: When and How Hidden Biases Transfer
Simon Schrodi, Elias Kempf, Fazl Barez +1
Language models can transfer hidden biases during distillation. For example, a teacher that "likes owls" can make its student "like owls" too, even when the training data consists…
Simple LLM Baselines are Competitive for Model Diffing
Elias Kempf, Simon Schrodi, Bartosz CywiÅski +3
Standard LLM evaluations only test capabilities or dispositions that evaluators designed them for, missing unexpected differences such as behavioral shifts between model revisions…
When and How Does CLIP Enable Domain and Compositional Generalization?
Elias Kempf, Simon Schrodi, Max Argus +1
The remarkable generalization performance of contrastive vision-language models like CLIP is often attributed to the diversity of their training distributions. However, key questio…
Revisiting SMoE Language Models by Evaluating Inefficiencies with Task Specific Expert Pruning
Soumajyoti Sarkar, Leonard Lausen, Volkan Cevher +3
Sparse Mixture of Expert (SMoE) models have emerged as a scalable alternative to dense models in language modeling. These models use conditionally activated feedforward subnetworks…
Concept Bottleneck Models Without Predefined Concepts
Simon Schrodi, Julian Schur, Max Argus +1
There has been considerable recent interest in interpretable concept-based models such as Concept Bottleneck Models (CBMs), which first predict human-interpretable concepts and the…
Eureka-Moments in Transformers: Multi-Step Tasks Reveal Softmax Induced Optimization Problems
David T. Hoffmann, Simon Schrodi, Jelena BratuliÄ +3
In this work, we study rapid improvements of the training loss in transformers when being confronted with multi-step decision tasks. We found that transformers struggle to learn th…