UTokyo/RIKEN lab brings CAROT cross-lingual alignment method to EMNLP, +11.2 accuracy points
Bollegala · x · 2026-10-24
The UTokyo/RIKEN Yanaka lab will present two main conference papers and one WMT paper at EMNLP 2026. CAROT aligns LLM representations across languages via token-level optimal transport in a language-agnostic space while explicitly preserving language-specific representations. Inference-time steering improves multilingual accuracy by up to 11.2 points, and using CAROT representations as training targets beats existing CLA methods in 11 of 18 evaluation settings.
More from Research
- Free book: Laurent Younes' 'Introduction to Machine Learning' covers the math behind core ML — TheTuringPost · 2026-10-24
- EvalResearchBench tests whether AI agents can build their own evaluators for recursive self-improvement — AG2AI · 2026-10-24
- NeRF creator Ben Mildenhall teases real-time inverse rendering demo — XRarchitect · 2026-10-24
- Reuse, Don't Recompute: A 4 A.M. ICLR Story Meets Meta² Agent Evolution — furongh · 2026-10-24
- Mapping Economic Theory With Agentic AI: LLM Agents Turn Theory Papers Into Graphs — mattbeane · 2026-10-24
- Robotics model leaderboards are premature, argues practitioner: task-specific evals first — paigeinsf · 2026-10-24