About
Research & Development
MindVL: A New Multimodal AI Model Achieves High Performance on Ascend NPUs with Less Data
TLDR: MindVL is a new multimodal large language model (MLLM) trained on Huawei's Ascend NPUs. It uses native-resolution Vision Transformers to process images without...
Research & Development
Robots Learn to Multitask: A New Approach for Handling Multiple Urgent Questions in 3D Environments
TLDR: This paper introduces Embodied Questions Answering (EQsA), a new problem setting where embodied agents must handle multiple, asynchronous questions with varying urgencies in...
Research & Development
Unpacking the Role of Code Semantics in Large Language Models: A New Study’s Surprising Insights
TLDR: A new research paper investigates whether execution trace-based semantic information helps Code Large Language Models (LLMs) in understanding and generating code. The study...
Research & Development
CoachMe: An AI System for Detailed Movement Correction in Sports
TLDR: CoachMe is a novel AI model that generates precise, sport-specific coaching instructions by comparing a learner's motion to a reference. It analyzes temporal...
Research & Development
Accelerating Vision-Language Models with Adaptive Compression and Online Training
TLDR: SpecVLM is a new system that significantly speeds up Vision-Language Models (VLMs) by combining speculative decoding with an adaptive visual token compressor and...
Research & Development
Targeted Forgetting: A New Method to Safely Unlearn Harmful Knowledge in AI Models
TLDR: The research paper "COLLAPSE OF IRRELEVANT REPRESENTATIONS (CIR) ENSURES ROBUST AND NON-DISRUPTIVE LLM UNLEARNING" introduces a novel technique called CIR for effectively removing...