CLaS-Bench : A Cross-Lingual Alignment and Steering Benchmark
Daniil Gurgurov Yusser Al Ghussin Tanja Bäumel Cheng-Ting Chou Patrick Schramowski Marius Mosbach Josef van Genabith Simon Ostermann · 2026
Read CLaS-Bench : A Cross-Lingual Alignment and Steering Benchmark by Daniil Gurgurov Yusser Al Ghussin Tanja Bäumel Cheng-Ting Chou Patrick Schramowski Marius Mosbach Josef van Genabith Simon Ostermann (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-08331.txt.
Description
Daniil Gurgurov and co-authors release CLaS-Bench, a benchmark for cross-lingual alignment and steering. Their 2026 arXiv paper addresses how to control language model behavior in multilingual settings. Steering changes internal representations during inference, offering a more efficient and interpretable route than prompting or fine-tuning.
A benchmark is needed because steering claims are hard to compare across languages. CLaS-Bench gives shared tasks for alignment and steering tests. Builders of multilingual systems can check whether a steering method works beyond English. The work supports safer control of models that serve users in many languages at once.
Catalog record
| CREATOR | Daniil Gurgurov Yusser Al Ghussin Tanja Bäumel Cheng-Ting Chou Patrick Schramowski Marius Mosbach Josef van Genabith Simon Ostermann |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 67 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-08331