Cross-lingual transfer in LLMs is weaker than expected
willdepue · x · 2026-08-20
In a discussion about cross-lingual transfer, the commenter notes that while evident, the capability is significantly weaker than hoped for five years ago. Even simple cross-lingual transfer involving the same topics performs much worse than one would expect.
Related event: Discussion: LLM Cross-Lingual Transfer Weaker Than Expected(3 posts)→
More from Research
- Moderna personalized mRNA cancer vaccine succeeds in Phase 3 trial — thedealdirector · 2026-08-20
- AI Detectors Biased Against Non-Native Speakers? Pangram 4 Achieves Zero False Positives — TuhinChakr · 2026-08-20
- Cell paper asks: could productive human transcription factors exist that never evolved? — anshulkundaje · 2026-08-20
- CRUX 2 study finds frontier agents struggle with autonomous AI research judgment — random_walker · 2026-08-20
- Can public chat data predict real-world AI misalignments? — yoavartzi · 2026-08-20
- Research exposes LLM API vulnerability leaking hidden chain-of-thought — burkov · 2026-08-20