544 citations · 561 across the 16 of their papers we have counts for
5 papers · 1 filter
Focus on Likely Classes for Test-Time Prediction
Johannes Schneider
We ask: Can focusing on likely classes of a single, in-domain sample improve accuracy? Prior work argued "no", we answer "yes" on average. Standard entropy minimization yields larg…
Using Phonemes in cascaded S2S translation pipeline
Rene Pilz, Johannes Schneider
This paper explores the idea of using phonemes as a textual representation within a conventional multilingual simultaneous speech-to-speech translation pipeline, as opposed to the…
Causally Reliable Concept Bottleneck Models
Giovanni De Felice, Arianna Casanova Flores, Francesco De Santis +4
Concept-based models are an emerging paradigm in deep learning that constrains the inference process to operate through human-interpretable variables, facilitating explainability a…
What comes after transformers? -- A selective survey connecting ideas in deep learning
Johannes Schneider
Transformers have become the de-facto standard model in artificial intelligence since 2017 despite numerous shortcomings ranging from energy inefficiency to hallucinations. Researc…
Understanding and Leveraging the Learning Phases of Neural Networks
Johannes Schneider, Mohit Prabhushankar
The learning dynamics of deep neural networks are not well understood. The information bottleneck (IB) theory proclaimed separate fitting and compression phases. But they have sinc…