activity
20242026
collaborators

10 papers

cs.LG2026

Steering Recurrent Reasoners at Inference Time with Readout Feedback

Shunsuke Kamiya, Masanori Koyama, Seongcheol Jeong +5

Recurrent models, which repeatedly update latent states with shared computation blocks, have emerged as powerful architectures for solving complex reasoning tasks. Existing inferen…

cs.LG2026

Looped Transformers with Source-Centered State Evolution

Bum Jun Kim, Kohei Hayashi, Shunsuke Kamiya +3

Looped Transformers create a useful train- and test-time compute axis by reusing the same Transformer block over recurrent depth, increasing effective depth at a fixed parameter co…

cs.LG2026

Anti Mode-Collapse in Mean-Field Transformer via Auxiliary Variables

Masaaki Imaizumi, Masanori Koyama, Noboru Isobe +1

We use a mean-field-based transformer model to theoretically investigate how auxiliary variables, such as positional encoding, prevent mode collapse of self-attention mechanisms. T…

cs.LG2026

C-voting: Confidence-Based Test-Time Voting without Explicit Energy Functions

Kenji Kubo, Shunsuke Kamiya, Masanori Koyama +3

Neural network models with latent recurrent processing, where identical layers are recursively applied to the latent state, have gained attention as promising models for performing…

cs.LG2026

Exploration of Fast-Slow Latent Recurrence for Train-Short, Test-Long Generalization

Shota Takashiro, Masanori Koyama, Takeru Miyato +3

We study out of distribution generalization in streaming tasks where models are trained on short sequences but must operate over much longer, unknown horizons under bounded memory.…

cs.LG2025

DiffusionBlocks: Block-wise Neural Network Training via Diffusion Interpretation

Makoto Shing, Masanori Koyama, Takuya Akiba

End-to-end backpropagation requires storing activations throughout all layers, creating memory bottlenecks that limit model scalability. Existing block-wise training methods offer…