Center for Advanced Intelligence Project

[NeurIPS 2026]

Main Track

Oral

• Can LLM Agents Respond to Disasters? Benchmarking Heterogeneous Geospatial Reasoning in Emergency Operations
  Junjue Wang (The University of Tokyo / RIKEN AIP)
  Weihao Xuan (The University of Tokyo / RIKEN AIP)
  Heli Qi (RIKEN AIP)
  Pengyu Dai (The University of Tokyo / RIKEN AIP)
  Kunyi Liu (RIKEN AIP)
  Hongruixuan Chen (RIKEN AIP)
  Zhuo Zheng (Stanford University)
  Junshi Xia (RIKEN AIP)
  Stefano Ermon (Stanford University)
  Naoto Yokoya (The University of Tokyo / RIKEN AIP)

• From Average Sensitivity to Small-Loss Regret Bounds under Random-Order Model
  Shinsaku Sakaue (CyberAgent / National Institute of Informatics / RIKEN AIP)
  Yuichi Yoshida (National Institute of Informatics)

Spotlight

• Non-asymptotic Implicit Bias of Logistic Regression at Early-stage Gradient Descent Dynamics
  Han Bao (The Institute of Statistical Mathematics / Tohoku University / RIKEN AIP)

• Toward Minimal-dimensional Convex Calibrated Surrogate Losses for Classification with Rejection
  Yuzhou Cao (Nanyang Technological University)
  Han Bao (The Institute of Statistical Mathematics / Tohoku University / RIKEN AIP)
  Bo An (Nanyang Technological University)

• The Implicit Bias of Hyperbolic Representation Learning for Multiclass Data: A Busemann Risk Perspective
  Xingrun Li (The University of Tokyo)
  Sho Kuno (The University of Tokyo)
  Yusuke Mukuta (The University of Tokyo / RIKEN AIP)
  Xin Yang (The University of Tokyo)
  Tatsuya Harada (The University of Tokyo / RIKEN AIP)

• Post-ADC Inference: Valid Inference After Active Data Collection
  Shuichi Nishino (Nagoya University / RIKEN AIP)
  Tomohiro Shiraishi (Nagoya University / RIKEN AIP)
  Teruyuki Katsuoka (Nagoya University)
  Ichiro Takeuchi (Nagoya University / RIKEN AIP)

Poster

• Estimating and Orthogonalizing Unknown Pre-training Gradients for Continual Fine-tuning of Large Language Models
  Bing Wang (Jilin University)
  Changchun Li (Jilin University)
  Xin-Qiang Cai (RIKEN AIP)
  Lin Yuanbo Wu (University of Warwick)
  Ximing Li (Jilin University / RIKEN AIP)
  Gang Niu (RIKEN AIP)
  Masashi Sugiyama (RIKEN AIP / The University of Tokyo)
• Rethinking Learning from Label Proportions via Moment Matching
  Tianhao Ma (The University of Tokyo)
  Wei Wang (RIKEN AIP)
  Yivan Zhang (The University of Tokyo / RIKEN AIP)
  Dong-Dong Wu (The University of Tokyo / RIKEN AIP)
  Takashi Ishida (RIKEN AIP / The University of Tokyo)
  Gang Niu (RIKEN AIP)
  Masashi Sugiyama (RIKEN AIP / The University of Tokyo)
• Do Coding Agents Deceive Us? Detecting and Preventing Cheating via Capped Evaluation with Randomized Tests
  Thanawat Lodkaew (The University of Tokyo / RIKEN AIP)
  Johannes Ackermann (The University of Tokyo / RIKEN AIP)
  Soichiro Nishimori (The University of Tokyo / RIKEN AIP)
  Nontawat Charoenphakdee (Preferred Networks / RIKEN AIP)
  Masashi Sugiyama (RIKEN AIP / The University of Tokyo)
  Takashi Ishida (RIKEN AIP / The University of Tokyo)
• Watermarking as a Learned Intrinsic Property of Diffusion Models
  Xiang Li (George Mason University)
  Dong-Dong Wu (The University of Tokyo / RIKEN AIP)
  Masashi Sugiyama (RIKEN AIP / The University of Tokyo)
  Lannan Luo (George Mason University)
  Qiang Zeng (George Mason University)
• DELTA: Robustly Training Label-Conditional Diffusion Models with Weak Annotations
  Dong-Dong Wu (The University of Tokyo / RIKEN AIP)
  Jiacheng Cui (Mohamed bin Zayed University of Artificial Intelligence)
  Wei Wang (RIKEN AIP)
  Zhiqiang Shen (Mohamed bin Zayed University of Artificial Intelligence)
  Masashi Sugiyama (RIKEN AIP / The University of Tokyo)
• Geometry-Aware Directional Alignment for Coherent Model Merging
  Zhikang Chen (University of Oxford)
  Sen Cui (Tsinghua University)
  Deheng Ye (Tencent)
  Min Zhang (East China Normal University)
  Gang Niu (RIKEN AIP)
  Yu Zhang (Southern University of Science and Technology)
  Masashi Sugiyama (RIKEN AIP / The University of Tokyo)
  Tingting Zhu (University of Oxford)
• Bayes-Optimal Balanced Error Rate and Area Under the Receiver Operating Characteristic Curve: Estimation and Evaluation of Estimators
  Ryota Ushio (The University of Tokyo / RIKEN AIP)
  Takashi Ishida (RIKEN AIP / The University of Tokyo)
  Masashi Sugiyama (RIKEN AIP / The University of Tokyo)
• Dataset Mismatch Matters in Group Relative Policy Optimization for Reinforcement Learning from Verifiable Rewards
  Zhen-Yu Zhang (RIKEN AIP)
  Jiandong Zhang (Northeastern University)
  Huaxiu Yao (University of North Carolina at Chapel Hill)
  Gang Niu (RIKEN AIP)
  Masashi Sugiyama (RIKEN AIP / The University of Tokyo)
• Asymptotically Exact Negative Guidance of Diffusion Models via Positive-Unlabeled Learning
  Hirohane Takagi (The University of Tokyo)
  Masashi Sugiyama (RIKEN AIP / The University of Tokyo)
• When and Why Does Multi-Agent Debate Fail and Does It Really Underperform?
  Yongqiang Chen (MBZUAI / Carnegie Mellon University)
  Gang Niu (RIKEN AIP)
  James Cheng (Chinese University of Hong Kong)
  Bo Han (Hong Kong Baptist University / RIKEN AIP)
  Masashi Sugiyama (RIKEN AIP / The University of Tokyo)
• VI-CuRL: Stabilizing Verifier-Independent Reinforcement Learning Reasoning via Confidence-Guided Variance Reduction
  Xin-Qiang Cai (RIKEN AIP)
  Masashi Sugiyama (RIKEN AIP / The University of Tokyo)
• Practical Estimation of the Bayes Optimal Fairness-Accuracy Tradeoff with Soft Labels
  Mohit Sharma (IIIT Delhi / RIKEN AIP)
  Okan Koç (RIKEN AIP)
  Amit Deshpande (Microsoft Research)
  Takashi Ishida (RIKEN AIP / The University of Tokyo)
  Gang Niu (RIKEN AIP)
  Rajiv Ratn Shah (IIT Kanpur / IIIT Delhi)
  Masashi Sugiyama (RIKEN AIP / The University of Tokyo)
• Unified Approach for Weakly Supervised Multicalibration
  Futoshi Futami (The University of Osaka / The University of Tokyo / RIKEN AIP)
  Takashi Ishida (The University of Tokyo / RIKEN AIP)
• UnlearningSoup: Is Repeated Tuning Necessary for Large Language Model Unlearning?
  Puning Yang (Mohamed bin Zayed University of Artificial Intelligence)
  Qizhou Wang (Hong Kong Baptist University / RIKEN AIP)
  Junchi Yu (University of Oxford)
  Bo Han (Hong Kong Baptist University / RIKEN AIP)
  Xiuying Chen (Mohamed bin Zayed University of Artificial Intelligence)
• On the Fragility of Latent Knowledge: Layer-wise Influence under Unlearning in Large Language Model
  Jianing Zhu (University of Texas at Austin)
  Zongze Li (Hong Kong Baptist University)
  Chandler Squires (Carnegie Mellon University)
  Qizhou Wang (Hong Kong Baptist University / RIKEN AIP)
  Bo Han (Hong Kong Baptist University / RIKEN AIP)
  Pradeep Ravikumar (Carnegie Mellon University)
• Simple Projection-Free Algorithm for Contextual Recommendation with Logarithmic Regret and Robustness
  Shinsaku Sakaue (CyberAgent / National Institute of Informatics / RIKEN AIP)
• Flow Matching from Viewpoint of Proximal Operators
  Kenji Fukumizu (The Institute of Statistical Mathematics)
  Wei Huang (RIKEN AIP / The Institute of Statistical Mathematics)
  Han Bao (The Institute of Statistical Mathematics / Tohoku University / RIKEN AIP)
  Shuntuo Xu (East China Normal University)
  Nisha Chandramoorthy (University of Chicago)
• Denoise First, Orthogonalize Later: Understanding Momentum in Muon via Spectral Filtering
  Xianliang Li (The Institute of Statistical Mathematics / SOKENDAI)
  Zihan Zhang (National Institute of Informatics / SOKENDAI)
  Weiyang Liu (The Chinese University of Hong Kong)
  Han Bao (The Institute of Statistical Mathematics / Tohoku University / RIKEN AIP)
• Tight Gap-Dependent Regret Bounds and Problem-Independent Bounds for Cost-aware Cascading Bandits
  Yuji Tamakoshi (The University of Tokyo / RIKEN AIP)
  Shinji Ito (The University of Tokyo / RIKEN AIP)
• Posterior-Tracking for Best-Arm Identification in Bernoulli Bandits
  Kaito Ariu (CyberAgent, Inc.)
  Junpei Komiyama (MBZUAI / RIKEN AIP)
• Reliable Chain-of-Thought via Prefix Consistency
  Naoto Iwase (Nagoya University)
  Yuki Ichihara (NAIST / MBZUAI)
  Mohammad Atif Quamar (MBZUAI)
  Junpei Komiyama (MBZUAI / RIKEN AIP)
• CITE: Anytime-Valid Statistical Inference in LLM Self-Consistency
  Hirofumi Ota (The University of Tokyo)
  Naoto Iwase (Nagoya University)
  Yuki Ichihara (NAIST / MBZUAI)
  Junpei Komiyama (MBZUAI / RIKEN AIP)
  Masaaki Imaizumi (The University of Tokyo)
• An Efficient Algorithm for Thresholding Monte Carlo Tree Search
  Shoma Nameki (Hokkaido University)
  Atsuyoshi Nakamura (Hokkaido University)
  Junpei Komiyama (MBZUAI / RIKEN AIP)
  Koji Tabata (Hokkaido University)
• Policy Optimization Achieves Data-Dependent Regret Bounds in MDPs with Unknown Transitions
  Mingyi Li (The University of Tokyo)
  Taira Tsuchiya (The University of Tokyo / RIKEN AIP)
  Kenji Yamanishi (The University of Tokyo)
• Near-Optimal Best-of-Both-Worlds Algorithms for Decoupled Exploration and Exploitation in Multi-armed Bandits
  Hibiki Sekiya (The University of Tokyo)
  Shinji Ito (The University of Tokyo / RIKEN AIP)
• Curvature-Dependent Lower Bounds for Riemannian Online Convex Optimization
  Hibiki Fukushima (The University of Tokyo / RIKEN AIP)
  Shinji Ito (The University of Tokyo / RIKEN AIP)
• Reward-Estimated Hypergradient for Bilevel Reinforcement Learning with Black-Box Follower
  Shigeki Kusaka (University of Tsukuba / RIKEN AIP / Institute of Science Tokyo)
  Mikoto Kudo (University of Tsukuba / RIKEN AIP)
  Takumi Tanabe (LY Corporation)
  Akifumi Wachi (LY Corporation)
  Youhei Akimoto (University of Tsukuba / RIKEN AIP / Institute of Science Tokyo)
• Guided Data Generation for Understanding Model Behavior
  E. Mehmet Kıral (RIKEN AIP)
  Nurşen Aydın (Warwick Business School)
  İ. Ülker Birbil (University of Amsterdam)
• Continual Learning in Modern Hopfield Networks with an Application to Diffusion Models
  Ken Takeda (The University of Tokyo / AIST)
  Masafumi Oizumi (The University of Tokyo)
  Ryo Karakida (AIST / RIKEN AIP)
• Don’t Waste Population: Post-Anneal Refinement for Combinatorial Optimization
  Yuma Ichikawa (Fujitsu Limited / RIKEN AIP)

• A Trust Region Approach for Learning Schrödinger Bridges
  Takeshi Koshizuka (RIKEN AIP / Kyushu University)
  Denis Blessing (Karlsruhe Institute of Technology)
  Max Zimmer (Zuse Institute Berlin / Technical University of Berlin)
  Sebastian Pokutta (Zuse Institute Berlin / Technical University of Berlin)
  Lorenz Richter (dida / Zuse Institute Berlin)
• Symplectic Neural Operators for Learning Infinite Dimensional Hamiltonian Systems
  Makara Yeang (Kobe University / RIKEN AIP)
  Yusuke Tanaka (NTT)
  Takashi Matsubara (Hokkaido University)
  Takaharu Yaguchi (Kyushu University / RIKEN AIP)
• Approximation of Maximally Monotone Operators: A Graph Convergence Perspective
  Takashi Furuya (Doshisha University / RIKEN AIP)
  Yury Korolev (University of Bath)
  Takaharu Yaguchi (Kyushu University / RIKEN AIP)
• Importance-Weighted Operator Learning Under Probability Measure Shifts
  Lei Sun (NAIST)
  Yusuke Tanaka (NTT)
  Xiaocheng Shang (University of Birmingham)
  Takaharu Yaguchi (Kyushu University / RIKEN AIP)
  Tomoharu Iwata (NTT)
• Function Graph Transformers Universally Approximate Operators Between Function Spaces
  Takashi Furuya (Doshisha University / RIKEN AIP)
  S. David Mis (Rice University)
  Ivan Dokmanić (University of Basel)
  Maarten V. de Hoop (Rice University)
  Matti Lassas (University of Helsinki)
• RAM-H1200: A Unified Evaluation and Dataset on Hand Radiographs for Rheumatoid Arthritis
  Songxiao Yang (Institute of Science Tokyo)
  Haolin Wang (Hokkaido University)
  Yao Fu (Hokkaido University)
  Junmu Peng (Hokkaido University)
  Lin Fan (Southwest Jiaotong University)
  Hongruixuan Chen (RIKEN AIP / The University of Tokyo)
  Jian Song (RIKEN AIP)
  Masayuki Ikebe (Hokkaido University)
  Shinya Takamaeda-Yamazaki (The University of Tokyo / RIKEN AIP)
  Masatoshi Okutomi (Institute of Science Tokyo)
  Tamotsu Kamishima (Hokkaido University)
  Yafei Ou (RIKEN AIP)
• OFBD: Object-Focused Background Debiasing for Long-Tailed Learning
  Shenghan Chen (Shandong University)
  Yiming Liu (Shandong University of Science and Technology)
  Zhipeng Deng (Westlake University)
  Haolin Wang (Hokkaido University)
  Jiale Zhou (Westlake University)
  Zhijian Wu (Westlake University)
  Xiankai Lu (Shandong University)
  Yafei Ou (RIKEN AIP)
  Yefeng Zheng (Westlake University)
• Generalization Error Bounds for Picard-Type Operator Learning in Nonlinear Parabolic PDEs
  Koichi Taniguchi (Shizuoka University)
  Sho Sonoda (RIKEN AIP / CyberAgent)
• Winning Lottery Tickets in Neural Networks via a Quantum-Inspired Classical Algorithm
  Natsuto Isogai (The University of Tokyo)
  Hayata Yamasaki (The University of Tokyo)
  Sho Sonoda (RIKEN AIP / CyberAgent)
  Mio Murao (The University of Tokyo)
• Mirror Descent-Ascent for Mean-Field Min-Max Problems
  Razvan-Andrei Lascu (RIKEN AIP)
  Mateusz B. Majka (Heriot-Watt University)
  Lukasz Szpruch (University of Edinburgh)
• PPO in the Fisher-Rao Geometry
  Razvan-Andrei Lascu (RIKEN AIP)
  David Siska (University of Edinburgh)
  Lukasz Szpruch (University of Edinburgh)
• Last-Iterate Convergence of Single-Loop Stochastic Methods for Constrained Convex-Concave Minimax Problems
  Taoli Zheng (RIKEN AIP)
  Jiajin Li (University of British Columbia)
  Anthony Man-Cho So (The Chinese University of Hong Kong)
• Mamba Can Learn Low-Dimensional Targets In-Context via Test-Time Feature Learning
  Junsoo Oh (KAIST)
  Wei Huang (RIKEN AIP / The Institute of Statistical Mathematics)
  Taiji Suzuki (The University of Tokyo / RIKEN AIP)
• Parallel Computation Algorithms and Convergence Guarantees for Mean-Field Langevin Dynamics
  Yoshihito Okamoto (The University of Tokyo)
  Huanjian Zhou (The University of Tokyo)
  Taiji Suzuki (The University of Tokyo / RIKEN AIP)
• Test Time Training Enhances In-Context Learning of Nonlinear Functions
  Kento Kuwataka (The University of Tokyo)
  Taiji Suzuki (The University of Tokyo / RIKEN AIP)
• How Neural Reward Models Learn Features for Policy Optimization: A Single-Index Analysis
  Rei Higuchi (The University of Tokyo / RIKEN AIP)
  Ryotaro Kawata (The University of Tokyo)
  Akifumi Wachi (LY Corporation)
  Shokichi Takakura (LY Corporation)
  Kohei Miyaguchi (LY Corporation)
  Taiji Suzuki (The University of Tokyo / RIKEN AIP)
• The Mechanism of Weak-to-Strong Generalization: Feature Elicitation from Latent Knowledge
  Ryoya Awano (The University of Tokyo)
  Taiji Suzuki (The University of Tokyo / RIKEN AIP)
• Learning Survival Models with Right-Censored Reporting Delays
  Yuta Shikuri (The Graduate University for Advanced Studies)
  Hironori Fujisawa (The Institute of Statistical Mathematics / The Graduate University for Advanced Studies / RIKEN AIP)
• Bayesian Causal Stress Testing: Posterior Fragility of Treatment-Effect Conclusions
  Makoto Nakakita (RIKEN AIP)
  Teruo Nakatsuma (Keio University)
• DriftWeight: Repulsive Drift in Mean-Flow Space for Neural Network Weight Generation
  Reo Iizuka (Boston Consulting Group)
  Hiroaki Shioya (Boston Consulting Group)
  Naoto Yokoya (The University of Tokyo / RIKEN AIP)
• Information-Theoretic Generalization Bounds for Sequential Decision Making
  Futoshi Futami (The University of Osaka / The University of Tokyo / RIKEN AIP)
  Masahiro Fujisawa (The University of Osaka / RIKEN AIP)
• Information-Theoretic Generalization for Set-Input Optimization-Valued Objectives
  Futoshi Futami (The University of Osaka / The University of Tokyo / RIKEN AIP)
  Masahiro Fujisawa (The University of Osaka / RIKEN AIP)
• Minimax Optimal Estimation of Transport-Growth Pairs in Unbalanced Optimal Transport
  D. Ponnoprat (Chiang Mai University)
  N. Isobe (RIKEN AIP)
  M. Imaizumi (The University of Tokyo / RIKEN AIP / Kyoto University)
• Extended Wasserstein-GAN Approach to Causal Distribution Learning: Density-Free Estimation and Minimax Optimality
  S. Tamano (The University of Tokyo)
  M. Imaizumi (The University of Tokyo / RIKEN AIP / Kyoto University)
• Spectrum-Adaptive Generalization Bounds for Trained Deep Transformers
  M. Sakai (The University of Tokyo / RIKEN AIP)
  M. Imaizumi (The University of Tokyo / RIKEN AIP / Kyoto University)
• Training-Induced Escape from Token Clustering in a Mean-Field Formulation of Transformers
  N. Isobe (RIKEN AIP)
  D. Inoue (Imperial College London)
  M. Imaizumi (The University of Tokyo / RIKEN AIP / Kyoto University)
• Optimal In-Context Learning of Autoregressive Processes under Heterogeneous Second-Order Moments of the Prompts
  H. Tseran (RIKEN AIP)
  M. Imaizumi (The University of Tokyo / RIKEN AIP / Kyoto University)
• Value-Rectified Distillation for Flow-based Offline Reinforcement Learning
  Ke Jiang (Nanjing University of Aeronautics and Astronautics)
  Wen Jiang (Nanjing University of Aeronautics and Astronautics)
  Yoshinobu Kawahara (The University of Osaka / RIKEN AIP)
  Xiaoyang Tan (Nanjing University of Aeronautics and Astronautics)
• Causal Discovery via Transformed Low-Rank Quantile Surfaces
  Ryo Kamimura (Shiga University / RIKEN AIP)
  Thong Pham (The University of Osaka / Shiga University / RIKEN AIP)
• Beyond ICA: Identifiability by Symmetry Breaking
  Pengzhou Wu (The University of Osaka / RIKEN AIP)
• Structured Unitary Tensor Network Representations for Circuit-Efficient Quantum Data Encoding
  Guang Lin (RIKEN AIP)
  Toshihisa Tanaka (Tokyo University of Agriculture and Technology)
  Qibin Zhao (RIKEN AIP)
• Score-based Variational Inference via Quantum Maximally Mixed States
  Yuchen Cong (Juntendo University / RIKEN AIP)
  Zerui Tao (RIKEN AIP)
  Chao Li (RIKEN AIP)
  Zhe Sun (Juntendo University)
  Qibin Zhao (RIKEN AIP)
• NoTA: Normalized Tensor Adaptation for Parameter-Efficient Continual Learning
  Yusong Deng (Guangdong University of Technology)
  Yuning Qiu (RIKEN AIP)
  Qibin Zhao (RIKEN AIP)
  Guoxu Zhou (Guangdong University of Technology)
• Beyond Exemplar Selection: Value-Aware Memory Allocation in Replay-Based Continual Learning
  Yuhang Li (Guangdong University of Technology)
  Guoxu Zhou (Guangdong University of Technology)
  Zhenhao Huang (Guangdong University of Technology)
  Yuning Qiu (RIKEN AIP)
  Qibin Zhao (RIKEN AIP)
• Randomized Kriging Believer For Parallel Bayesian Optimization With Regret Bounds
  Shuhei Sugiura (RIKEN AIP)
  Ichiro Takeuchi (Nagoya University / RIKEN AIP)
  Shion Takeno (Nagoya University)
• SOAP-Bubbles: Structured Weight Uncertainty for Neural Networks
  Adrian Robert Minut (Sapienza University of Rome)
  Nico Daheim (Technical University Darmstadt)
  Marco Miani (Technical University of Denmark)
  Mohammad Emtiyaz Khan (RIKEN AIP)
  Wu Lin (University of Central Florida)
  Thomas Möllenhoff (RIKEN AIP)
• The Bayesian Learning Rule Beyond KL Geometry
  Sophia Sklaviadis (Instituto Superior Técnico / University of Lisbon)
  Wu Lin (University of Central Florida)
  Mario A. T. Figueiredo (Instituto Superior Técnico / University of Lisbon)
  Andre Martins (Instituto Superior Técnico / University of Lisbon)
  Thomas Möllenhoff (RIKEN AIP)
  Mohammad Emtiyaz Khan (RIKEN AIP)
• Incorporating Neural Network Structure in the Bayesian Learning Rule
  Eiki Shimizu (The Institute of Statistical Mathematics / Institute of Science Tokyo)
  Mohammad Emtiyaz Khan (RIKEN AIP)
  Thomas Möllenhoff (RIKEN AIP)
• VocalGrad: Evaluating Acoustic Perception in Audio Language Models
  Koki Ryu (The University of Tokyo / RIKEN AIP)
  Hitomi Yanaka (The University of Tokyo / RIKEN AIP)

Education Track

• How Good Can a Model Get? The Bayes Error as Performance Limit, Cheating Alarm and Prevention
  Thanawat Lodkaew (The University of Tokyo / RIKEN AIP)
  Takashi Ishida (RIKEN AIP / The University of Tokyo)

Related Laboratories