About
Research & Development
Unveiling Physical Laws with Interpretable Neural Operators
TLDR: KANO is a new neural operator that combines spectral and spatial representations to model complex, position-dependent physical dynamics. It outperforms traditional methods like...
Research & Development
Dynamic Expert Specialization: A New Approach for Multi-Domain AI Models
TLDR: DES-MoE is a framework for adapting Mixture-of-Experts (MoE) models to multiple domains without catastrophic forgetting. It uses an adaptive router, domain-guided expert specialization,...
Research & Development
Unveiling Clarity: A Novel Framework for Image Restoration in Challenging Weather
TLDR: LCDiff is a new framework for adverse weather image restoration that uses a Lumina-Chroma Decomposition Network (LCDN) in YCbCr color space to separate...
Research & Development
Bridging Text and Vision: A New Framework for Robot Affordance Learning
TLDR: This research introduces an informative framework for text-guided visual affordance learning, enabling robots to better understand object interactions. It addresses limitations of previous...
Research & Development
LifeAlign: Enabling Large Language Models to Evolve Without Forgetting
TLDR: LifeAlign is a new framework for Large Language Models (LLMs) that solves the problem of "catastrophic forgetting" during continuous learning of human preferences....
Research & Development
Tracking the Internal Development of Language Models
TLDR: This research uses "crosscoders" to track how interpretable features (concepts) emerge and evolve within large language models during pre-training. It identifies two phases:...