ACL 2026
SAC Highlight Award
Taiga Someya
Taiga Someya is a Ph.D. student at the University of Tokyo and a MATS 11.0 Fellow in the Redwood Research stream. His research spans natural language processing and sports analytics using deep learning. Former soccer U-15/16 Japan national team candidate and Kashiwa Reysol U-18 player.
Selected research
View all publicationsLessWrong, 2026
Models Trained to Evade Topic Probes Struggle to Answer Questions about Those Topics
ACL 2025
Oral + Panel (Top 0.8%)Information Locality as an Inductive Bias for Neural Language Models
LREC-COLING 2024
Best Paper Award (1/1554; Top 0.06%)Targeted Syntactic Evaluations on the Chomsky Hierarchy
EACL 2023 Findings
JBLiMP: Japanese Benchmark of Linguistic Minimal Pairs
Experience
View full CV- Research Fellow, MATS Research, Redwood Research streamSep 2026 – Present
- Visiting Scholar, New York University, Center for Data ScienceNov 2025 – May 2026
- Research Visitor, ETH Zürich, Department of Computer ScienceNov 2024 – Jun 2025
- Technical Staff, National Institute of Advanced Industrial Science and Technology (AIST)Nov 2022 – Mar 2026
Latest news
View all newsResearch post published on LessWrong
- Models Trained to Evade Topic Probes Struggle to Answer Questions about Those Topics
Two papers accepted to NeurIPS 2026 ED Track and IEEE GCCE 2026
- Global PIQA: Evaluating Commonsense Reasoning Across 100+ Languages and Cultures
- Inferring Collective Reward Preferences from Multi-Agent Trajectories with Gradient-Based Inverse Reinforcement Learning in Soccer
Joined MATS Research as a MATS Fellow
