UvA-MT
The machine translation research group at the University of Amsterdam, dedicated to research on frontier machine translation methods and participating in international evaluations such as WMT.
Timeline (last 90 days)
UvA-MT submitted an unconstrained-track solution covering all 23 translation directions in the WMT26 general machine translation shared task.
UvA-MT's WMT26 solution is entirely based on LLM inference-time optimization, without additionally training models.
The first core component of the UvA-MT solution is sequential sampling with refined prompts.
The second core component of the UvA-MT solution is LLM pairwise judging combined with circular voting to select the best from the candidate pool.
The paper does not disclose specific computational costs in the abstract.
All Facts (5)
UvA-MT's WMT26 solution is entirely based on LLM inference-time optimization, without additionally training models.
50%UnverifiedThe first core component of the UvA-MT solution is sequential sampling with refined prompts.
50%UnverifiedThe second core component of the UvA-MT solution is LLM pairwise judging combined with circular voting to select the best from the candidate pool.
50%UnverifiedThe paper does not disclose specific computational costs in the abstract.
50%UnverifiedUvA-MT submitted an unconstrained-track solution covering all 23 translation directions in the WMT26 general machine translation shared task.
50%