Pairwise Judging
A candidate selection strategy for LLM inference that pairs candidates and has the LLM judge their relative merits, then selects the best by tallying wins through round-robin voting. It is more stable and robust than absolute scoring.
Timeline (last 90 days)
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.
Offline evaluation by the research team on the WMT25 benchmark showed that pairwise-judging circular voting yields a very clear improvement over pointwise best-of-N selection.
All Facts (2)
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.
50%UnverifiedOffline evaluation by the research team on the WMT25 benchmark showed that pairwise-judging circular voting yields a very clear improvement over pointwise best-of-N selection.
50%