Research Blog

The Joy of Multiple-Choice

Box-ticking exams save ink and time.

Read post

OpenAI's Speculative Decoding,
Reverse-Engineered

Why LLMs are faster if we give them a draft to complete.

Read post

An Encoder Model for Swiss German

SwissBERT can now process written Swiss German.

Read post

Wenn ChatGPT den Smartvote-Fragebogen ausfüllt

Sind Sprachmodelle politisch voreingenommen?

Read post

Introducing SwissBERT

The multilingual language model for Switzerland.

Read post

Translation Puzzles are In‑context Learning Tasks

Large language models can tackle some hard linguistic tasks.

Read post

Three Diffusion Digressions

The Stable Diffusion release inspired me to make tiny concept art.

Read post

Lost and Found in Translation

Hypothetical reasoning can detect overtranslations and undertranslations.

Read post

NMTScore: Text Similarity via Translation

Multilingual translation models offer surprising ways of comparing two sentences.

Read post

The Limits of Minimal Sentence Pairs

Forced decisions between sentences are not always predictive of generated language.

Read post

When MT Distillation Leads to Bias

Distilled translation models tend to overgeneralize.

Read post

Evaluating Black-Box MT with Contrastive Conditioning

Why not use contrastive sources instead of contrastive translations?

Read post

More General Stance Detection with x-Stance

Introducing a dataset for multilingual and multi-target stance detection.

Read post

BERT for NER

How to apply BERT to the task of named entity recognition.

Read post

How NLP affects Gender Equality

A brief essay discussing problems for gender equality in my field of study.

Read post