About
Research & Development
MoKA: Enhancing LLM Adaptation with Gated Kronecker Mixtures
TLDR: MoKA (Mixture of Kronecker Adapters) is a new parameter-efficient fine-tuning (PEFT) method for large language models (LLMs). It overcomes limitations of traditional low-rank...
Research & Development
EmoSteer-TTS: Precise Emotion Control in Synthesized Speech Without Retraining
TLDR: EmoSteer-TTS is a new training-free method for fine-grained emotion control in text-to-speech (TTS) systems. It works by "steering" internal model activations to convert,...
Research & Development
Unlocking Plant Communication for Advanced Farming
TLDR: This research paper explores how understanding and engineering the communication within the phytobiome (plants, their environment, and associated organisms) can revolutionize smart agriculture....
Research & Development
AttZoom: Enhancing CNN Performance Through Focused Attention
TLDR: AttZoom is a new, easy-to-integrate spatial attention layer for convolutional neural networks (CNNs) that improves image feature extraction. It works by spatially emphasizing...
Research & Development
Assessing ChatGPT’s Live Video Assistance for Visually Impaired Individuals
TLDR: A study explored how blind or visually impaired (BVI) individuals use ChatGPT's live video chat for real-world assistance. While effective for static visual...
Research & Development
Boosting Japanese AI Reasoning with Vector Transfer
TLDR: A novel method called 'reasoning vectors' is introduced to enhance the reasoning capabilities of Japanese Large Language Models (LLMs) without extensive additional training...