Momose Oyama of the Sound Scene Understanding Team, and Hidetoshi Shimodaira, Visiting Scientist of the Sound Scene Understanding Team and the Causal Inference Team (Professor, Kyoto University), have received the Outstanding Paper Award at ACL 2025.
This study, which maps more than 1,000 language models into a probability-distribution space using log-likelihood vectors and quantifies the characteristics of their text generation, was highly evaluated by the conference.
- Paper Title: Mapping 1,000+ Language Models via the Log-Likelihood Vector
- Authors:
Momose Oyama (Kyoto University / RIKEN AIP)
Hiroaki Yamagiwa (Kyoto University)
Yusuke Takase (Kyoto University)
Hidetoshi Shimodaira (Kyoto University / RIKEN AIP)
For details on the research, please refer to the ACL2025 website:
https://aclanthology.org/2025.acl-long.1584/
In addition, an online lecture on the award-winning work is scheduled as part of the NLP Colloquium. For further information, please see the NLP Colloquium website.
https://nlp-colloquium-jp.github.io/schedule/2025-08-27_hidetoshi_shimodaira/
Date/Time: Wednesday, August 27, 2025, from 12:00 (approx. 20 minutes)
Title: A Map of Language Models: Visualizing Similarity through Probability Distributions and Information Geometry
Language: Japanese
