Two papers on token-level OT multilingual alignment and LVLM illusion perception accepted to EMNLP 2026
Bollegala · x · 2026-08-21
A researcher announced two papers accepted to EMNLP 2026 main: one improves LLM multilingual ability via token-level Optimal Transport for cross-lingual alignment; the other uses bistable illusions to compare perception mechanisms between large vision-language models (LVLMs) and humans. Papers coming to arXiv soon.
Related event: Two NLP Papers on Multilingual Alignment and Bias Accepted to EMNLP 2026(2 posts)→
More from Research
- Study: Larger models tolerate more data repetition during pretraining — StanfordAILab · 2026-08-22
- η-learning framework predicts rare events without observed extreme data — bravo_abad · 2026-08-22
- Anthropic and Muni both released AI agent protein binder designs — here's a comparison — anshulkundaje · 2026-08-22
- How AI helps pick immune targets in a promising mRNA cancer vaccine — HarperSCarroll · 2026-08-22
- Analysis of limitations in OpenAI's chain-of-thought monitorability evals — sarahwiegreffe · 2026-08-22
- Equilibrium Forcing Enables Adaptive Video Generation Without Noise Conditioning — kwangmoo_yi · 2026-08-22