Dream-Cubed
AI & ML interests
We are a Tokyo-based R&D company on a quest to create a new kind of foundational AI model based on nature-inspired intelligence.
Recent Activity
Sparser, Faster, Lighter Transformers
Sparser, Faster, Lighter Transformer Language Models
Continuous Thought Machines
because thought takes time and reasoning is a process.
TinySwallow
Compact Japanese models trained with "TAID: Temporally Adaptive Interpolated Distillation for Efficient Knowledge Transfer in Language Models"
KAME
KAME
DroPE
Extending the Context of Pretrained LLMs by Dropping Their Positional Embedding (https://www.arxiv.org/abs/2512.12167)
Reinforcement Learning Teachers
Students distilled from a 7B Reinforcement-Learned Teacher (RLT) from the paper "Reinforcement Learning Teachers of Test Time Scaling."
CycleQD
Dream-Cubed
KAME
KAME
Sparser, Faster, Lighter Transformers
Sparser, Faster, Lighter Transformer Language Models
DroPE
Extending the Context of Pretrained LLMs by Dropping Their Positional Embedding (https://www.arxiv.org/abs/2512.12167)
Continuous Thought Machines
because thought takes time and reasoning is a process.
Reinforcement Learning Teachers
Students distilled from a 7B Reinforcement-Learned Teacher (RLT) from the paper "Reinforcement Learning Teachers of Test Time Scaling."
TinySwallow
Compact Japanese models trained with "TAID: Temporally Adaptive Interpolated Distillation for Efficient Knowledge Transfer in Language Models"
CycleQD