9 papers
Shachi: A Modular, Controllable Framework for LLM-Based Agent-Based Modeling of Emergent Collective Behavior
So Kuroki, Yingtao Tian, Kou Misaki +3
How collective behaviors emerge from the interactions of individual LLM-driven agents is a central question in artificial life, yet controlled study of these emergent dynamics has…
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…
Feedback-to-Rubrics: Can We Learn Expert Criteria from Inline Comments?
Kotaro Yoshida, So Kuroki, Yuki Imajuku +4
Large language models (LLMs) are increasingly used for writing and review support, but their usefulness depends on context-dependent criteria, such as expert preferences or organiz…
KAME: Tandem Architecture for Enhancing Knowledge in Real-Time Speech-to-Speech Conversational AI
So Kuroki, Yotaro Kubo, Takuya Akiba +1
Real-time speech-to-speech (S2S) models excel at generating natural, low-latency conversational responses but often lack deep knowledge and semantic understanding. Conversely, casc…
Wider or Deeper? Scaling LLM Inference-Time Compute with Adaptive Branching Tree Search
Yuichi Inoue, Kou Misaki, Yuki Imajuku +3
Recent advances demonstrate that increasing inference-time computation can significantly boost the reasoning capabilities of large language models (LLMs). Although repeated samplin…
Drop-Upcycling: Training Sparse Mixture of Experts with Partial Re-initialization
Taishi Nakamura, Takuya Akiba, Kazuki Fujii +3
The Mixture of Experts (MoE) architecture reduces the training and inference cost significantly compared to a dense model of equivalent capacity. Upcycling is an approach that init…