News

9/27/2026

September 2026

Hanato Kikuchi’s paper, “Learning Large-Scale Modular Addition with an Auxiliary Modulus,” was accepted at NeurIPS 2026. This is joint work with Ryosuke Masuya, Kazuhiko Kawamoto, and Hiroshi Kera.
Alumnus Yusuke Marumo received the 2026 IPSJ Yamashita SIG Research Award for “Matching Non-Identical Objects,” research conducted while he was a student in our lab.

8/31/2026

August 2026

Toshinori Yamauchi received the Student Encouragement Award at the 29th Meeting on Image Recognition and Understanding (MIRU 2026) for “Zero-Shot Faithful Textual Explanations via Directional-Derivative Influence on Predictions.” The work also received the Audience Award. This is joint work with Kazuhiko Kawamoto (Chiba University) and Hiroshi Kera.
Hiroto Sato received the Student Encouragement Award at MIRU 2026 for “視覚言語モデルの命名困難な特徴に対する属性ヒントを用いた自動命名” (automatic naming of hard-to-name features in vision-language models using attribute hints). The work also received the Audience Award. This is joint work with Toshinori Yamauchi, Hiroshi Kera, and Kazuhiko Kawamoto (Chiba University).
We released Ryo Kamiya’s preprint, “What Does Prompt Learning Change? -A Natural-Language Concept Analysis of Vision-Language Models.” This is joint work with Kazuhiko Kawamoto (Chiba University).

7/31/2026

July 2026

Kera gave an invited talk, “Computational Algebra with Transformers: What Deep Learning Adds to Computational Algebra,” at Symbolic Computation and Machine Learning in Austria.
Kera received the Silver Reviewer Award at ICML 2026.

6/30/2026

June 2026

Hanato Kikuchi’s paper, “Learning Large-Scale Modular Addition with an Auxiliary Modulus,” was accepted at the ICML 2026 AI for Math Workshop. This is joint work with Kazuhiko Kawamoto (Chiba University).
A paper co-authored with Sai Ganesh Nagarajan (University of Southern Denmark), “Seeing Through Distractions: Stable Attribution via the Core,” was accepted at the ICML 2026 Workshop.
Toshinori Yamauchi’s paper, “Explaining Object Detectors via Collective Contribution of Pixels,” was accepted at CVPR 2026 and selected as a Highlight. This is joint work with Kazuhiko Kawamoto (Chiba University).

5/31/2026

May 2026

Soichiro Kumano’s paper, “Adversarially Pretrained Transformers May Be Universally Robust In-Context Learners,” was accepted at ICLR 2026. This is joint work with Toshihiko Yamasaki (The University of Tokyo).
Ayumu Ueyama posted the preprint “Guided Diffusion Sampling for Precipitation Forecast Interventions.” This is joint work with Kazuhiko Kawamoto (Chiba University).

3/31/2026

March 2026

Yusuke Marumo’s paper, “Matching Semantically Similar Non-Identical Objects,” was accepted at WACV 2026. This is joint work with Kazuhiko Kawamoto (Chiba University), and Satomi Tanaka and Shigenobu Hirano (Ricoh).
A paper co-authored with Yuya Kanazawa, Midori Tanaka, and Takahiko Horiuchi, “Leveraging Vision–Language Models for Semantic Interpretation of Historical Paintings: A Case Study on Religiousness,” was accepted at IS&T Electronic Imaging.
Yuya Kanazawa received the Best Student Paper Award for the same paper at the Computer Vision and Image Analysis of Art (CVAA) Conference.

1/31/2026

January 2026

Yusuke Marumo received the CVIM Encouragement Award for the feature matching between similar-but-non-identical objects, jointly with Kazuhiko Kawamoto (Chiba University), Reimi Tanaka, and Shigenobu Hirano (Ricoh).

12/31/2025

December 2025

Kera’s paper, “Computational Algebra with Attention: Transformer Oracles for Border Basis Algorithms,” was accepted at NeurIPS 2025. This is joint work with Nico Pelleriti, Max Zimmer, and Sebastian Pokutta (Zuse Institute Berlin), and Yuki Ishihara (Nihon University).
Toshinori Yamauchi posted the preprint “Zero-Shot Faithful Textual Explanations via Directional-Derivative Influence on Predictions.” This is joint work with Kazuhiko Kawamoto (Chiba University).
Kera received a Top Review Award at NeurIPS 2025.

11/30/2025

November 2025

Shingo Ayabe received the Young Researcher Award at IWACIII 2025 for “Online Fine-tuning for Robustness in Offline Reinforcement Learning: Actor-Critic vs Decision Transformer,” jointly with Kazuhiko Kawamoto (Chiba University).

10/31/2025

October 2025

Ryo Kamiya’s paper, “Explaining Learned Prompts in Vision-Language Models,” was accepted at the International Symposium on Nonlinear Theory and Its Applications (NOLTA). This is joint work with Kazuhiko Kawamoto (Chiba University).
Shunsei Koshibuchi’s paper, “OFASD: One-Shot Federated Anisotropic Scaling Distillation,” was accepted at the ICCV 2025 LIMIT Workshop. This is joint work with Kazuhiko Kawamoto (Chiba University).
Shingo Ayabe posted the preprint “Adversarial Fine-tuning in Offline-to-Online Reinforcement Learning for Robust Robot Control.” This is joint work with Kazuhiko Kawamoto (Chiba University).

9/30/2025

September 2025

Kera gave an invited talk, “Symbolic Intelligence: Algebra by Hands, Computers, and AI,” at the workshop “Deepening and Applications of Computational Algebraic Geometry via Gröbner bases” in Kyoto.

8/31/2025

August 2025

Kera’s paper “CALT: A Library for Computer Algebra with Transformer” was accepted at ISSAC 2025 (Short Communication).
Kera received the MIRU Paper Evaluation Contribution Award.
Kazuki Yoda received the MIRU Student Encouragement Award and the Audience Award at the 28th Meeting on Image Recognition and Understanding (MIRU 2025) for “Characterizing Transformers through QR Code Decoding,” jointly with Kazuhiko Kawamoto (Chiba University).

7/31/2025

July 2025

Yuta Sato’s paper, “Chain of Thought in Order: Discovering Learning-Friendly Orders for Arithmetic,” was accepted at the ICML 2025 AI for Math Workshop. This is joint work with Kazuhiko Kawamoto (Chiba University).
Kazuki Yoda’s paper, “Learning Moderately Input-Sensitive Functions: A Case Study in QR Code Decoding,” was accepted at the same workshop. This is joint work with Kazuhiko Kawamoto (Chiba University).
Shingo Ayabe’s paper, “Robustness evaluation of offline reinforcement learning for robot control against action perturbations,” was accepted at International Journal of Advanced Robotic Systems. This is joint work with Kazuhiko Kawamoto (Chiba University).

6/30/2025

June 2025

Takuma Fukuda’s paper, “Adapter Merging with Centroid Prototype Mapping for Scalable Class-Incremental Learning,” was accepted at CVPR 2025. This is joint work with Kazuhiko Kawamoto (Chiba University).
Kera posted the preprint “Gradient-Weighted, Data-Driven Normalization for Approximate Border Bases.” This is joint work with Achim Kehrein (Rhine-Waal University of Applied Sciences).

5/31/2025

May 2025

Posted the preprint “Training on Plausible Counterfactuals Removes Spurious Correlations,” co-authored with Shpresim Sadiku, Kartikeya Chitranshi, and Sebastian Pokutta (Zuse Institute Berlin).
Posted the preprint “Cross-Model Transfer of Task Vectors via Few-Shot Orthogonal Alignment,” co-authored with Kazuhiko Kawamoto (Chiba University).
Shun Arakawa received the Excellence Award (Poster Presentation) at the 2025 Annual Conference of the Japanese Society for Artificial Intelligence (39th), jointly with Kazuhiko Kawamoto (Chiba University).

4/30/2025

April 2025

Kera joined Chiba University’s Institute for Advanced Academic Research as Associate Professor, marking the launch of mai₂ Lab.
Xiying Dong’s paper, “Vulnerability Verification in Robot Control Using Decision Transformer,” was accepted at Communications in Computer and Information Science. This is joint work with Kazuhiko Kawamoto (Chiba University).
Tsuyoshi Takano’s paper, “Robustness Verification of Decision Transformer with Varying Noise-Augmented Data Ratios in Atari Games,” was accepted at the same venue. This is joint work with Kazuhiko Kawamoto (Chiba University).