About
Research & Development
MambaLite-Micro Brings Advanced AI to Tiny Microcontrollers
TLDR: MambaLite-Micro is the first system to successfully deploy Mamba-based neural networks on resource-constrained microcontrollers (MCUs). It uses a C-based, runtime-free inference engine with...
Research & Development
OpenEgo: A New Dataset for Training Dexterous Robots with Egocentric Video
TLDR: OpenEgo is a large-scale, multimodal egocentric dataset (1107 hours, 290 tasks) for dexterous manipulation. It unifies 21-joint hand-pose annotations and provides intention-aligned, timestamped...
Research & Development
Video Summarization Gets Smarter with Language-Guided Object Interactions
TLDR: VideoGraph is a novel approach to video summarization that uses language queries to guide the selection of keyframes. It models videos as spatiotemporal...
Research & Development
ICON 2: A New Path to Efficient LLM Alignment with Self-Generated Data
TLDR: ICON 2 is a novel framework for aligning Large Language Models (LLMs) with human preferences. It addresses the high cost and inefficiency of...
Research & Development
Enhancing Medical AI: A Causal Approach to Handling Incomplete Multimodal Data
TLDR: This paper introduces Causal Debiasing (CaD), a new framework for medical multimodal representation learning that tackles two common biases: missingness bias (from non-random...
Research & Development
LM-Searcher: Unifying Neural Architecture Search Across Domains Using LLMs
TLDR: LM-Searcher is a novel framework that utilizes Large Language Models (LLMs) for Neural Architecture Search (NAS) across diverse domains without requiring extensive domain-specific...