Cohere Labs posts 6 WMT26 papers, finds small models are not yet polyglots
Cohere_Labs · x · 2026-10-27
Cohere and Cohere Labs announced six papers at the WMT conference during EMNLP 2026:
- Contrastive ESA: human evaluation of multiple translations at once
- Unlocking Reasoning Capability on Machine Translation in LLMs
- North Small Translate technical report
- Findings of the WMT26 General MT Shared Task: contrastive dynamic human evaluation at scale
- Findings of Automated Translation Quality Evaluation
- WMT26 Multilingual Instruction Shared Task: concluding "Small Models Are Not Yet Polyglots"
The author lists include MT stalwarts such as Tom Kocmi and Philipp Koehn.
More from Research
- eLife study: LLMs poorly capture how word order shapes sentence meaning in the human brain — AnnaCiaunica · 2026-10-27
- Lean–text proof mismatch means AI math claims need long human verification — srchvrs · 2026-10-27
- Why the exponential time hypothesis matters for AI reasoning complexity — srchvrs · 2026-10-27
- Ring-aperture VCSELs show best LI performance, argue researchers — jwt0625 · 2026-10-27
- Dev quips: if 3SAT constants are cosmic, P=NP wouldn't matter anyway — sytelus · 2026-10-27
- Bittensor subnet SN73 rebrands to Elenost, benchmarking perception in a simulated city — markjeffrey · 2026-10-27