• Koki Maeda and Naoaki Okazaki. JaWildText: A Benchmark for Vision-Language Models on Japanese Scene Text Understanding. In Proceedings of the 20th International Conference on Document Analysis and Recognition (ICDAR), 2026.

• Koki Maeda and Naoaki Okazaki. JaWildText: A Benchmark for Vision-Language Models on Japanese Scene Text Understanding. In Proceedings of the 20th International Conference on Document Analysis and Recognition (ICDAR), 2026.

• Masanari Oi, Mahiro Ukai, Masahiro Kaneko, Okazaki Naoaki, and Nakamasa Inoue. Autoregressive Direct Preference Optimization. In Forty-Third International Conference on Machine Learning (ICML), Seoul, South Korea, July 2026.
• Masanari Oi, Koki Maeda, Ryuto Koike, Daisuke Oba, Nakamasa Inoue, and Naoaki Okazaki. From Correspondence to Actions: Human-Like Multi-Image Spatial Reasoning in Multi-modal Large Language Models. In Forty-Third International Conference on Machine Learning (ICML), Seoul, South Korea, July 2026.
• Sora Miyamoto, Daisuke Oba, and Naoaki Okazaki. Aligning Tree-Search Policies with Fixed Token Budgets in Test-Time Scaling of LLMs. In Forty-Third International Conference on Machine Learning (ICML), Seoul, South Korea, July 2026.
• Sangwhan Moon, Daisuke Oba, Youmi Ma, Tatsuya Hiraoka, and Naoaki Okazaki. Beyond Perplexity: UTF-8 Validity in Byte-aware Language Models. In Forty-Third International Conference on Machine Learning (ICML), Seoul, South Korea, July 2026.

• Masanari Oi, Mahiro Ukai, Masahiro Kaneko, Okazaki Naoaki, and Nakamasa Inoue. Autoregressive Direct Preference Optimization. In Forty-Third International Conference on Machine Learning (ICML), Seoul, South Korea, July 2026.
• Masanari Oi, Koki Maeda, Ryuto Koike, Daisuke Oba, Nakamasa Inoue, and Naoaki Okazaki. From Correspondence to Actions: Human-Like Multi-Image Spatial Reasoning in Multi-modal Large Language Models. In Forty-Third International Conference on Machine Learning (ICML), Seoul, South Korea, July 2026.
• Sora Miyamoto, Daisuke Oba, and Naoaki Okazaki. Aligning Tree-Search Policies with Fixed Token Budgets in Test-Time Scaling of LLMs. In Forty-Third International Conference on Machine Learning (ICML), Seoul, South Korea, July 2026.
• Sangwhan Moon, Daisuke Oba, Youmi Ma, Tatsuya Hiraoka, and Naoaki Okazaki. Beyond Perplexity: UTF-8 Validity in Byte-aware Language Models. In Forty-Third International Conference on Machine Learning (ICML), Seoul, South Korea, July 2026.

The laboratory information session will be held at the following date, time, and location:

• Date & Time: 14:00, 14:30, 15:00, 15:30, 16:00, 16:30, April 18, 2026 (Sat)
• Duration: Approximately 30 minutes (you may enter and leave freely)
• Location: Room 604, 6F, West Bldg. 8E, Ookayama Campus, Institute of Science Tokyo

If you wish to participate, please apply for a ticket via the Peatix page for the graduate information session.

During the laboratory information session, Professor Okazaki will first introduce the laboratory, followed by a Q&A session with current lab members.
If you are interested in joining the lab, please also visit this page.
We look forward to speaking with you on the day!

研究室説明会は以下の日時・場所で実施します．

• 日時: 2026年4月18日（土）14:00, 14:30, 15:00, 15:30, 16:00, 16:30
• 所要時間: 約30分（途中入退場可）
• 場所: 東京科学大学 大岡山キャンパス 西8号館E棟 6階 604室

参加を希望される方は，大学院説明会のPeatixのチケットの申し込みをお願いします．

研究室説明会では，岡崎教授からの研究室紹介の後に，研究室所属学生との質疑応答の時間を取る予定です． 配属希望の方はこちらのページもご覧ください． 当日は皆さんとお話しできることを楽しみにしています！