About
Research & Development
Boosting Sparse Autoencoders for Enhanced Domain-Specific AI Interpretability
TLDR: Sparse Autoencoders (SAEs) help interpret Large Language Models (LLMs) but struggle with domain-specific concepts. This paper introduces SAE Boost, a novel residual learning...
Research & Development
Understanding Physical Challenges in Vision-and-Language Navigation
TLDR: A new platform, VLN-PE, has been developed to test language-guided robot navigation in physically realistic environments. It supports various robot types and environmental...
Research & Development
MUPAX: A Unified Approach to Explainable AI Across Diverse Data Types
TLDR: MUPAX (Multidimensional Problem–Agnostic eXplainable AI) is a novel, deterministic, and model-agnostic XAI technique with guaranteed convergence. It uses structured perturbation analysis to identify...
Research & Development
GraspGen: Advancing Robotic Grasping with Diffusion Models and Smart Training
TLDR: GraspGen is a new framework that significantly improves robotic grasping by using a diffusion-based generative model to propose 6-DOF grasps and an efficient...
Research & Development
Evaluating Reasoning in Large Language Models with HATS: A New Hindi Analogy Test Set
TLDR: HATS introduces a new Hindi Analogy Test Set (405 multiple-choice questions from Indian government exams) to evaluate reasoning in multilingual Large Language Models...
Research & Development
Enhancing Vision Transformer Adaptation with Approximately Orthogonal Fine-Tuning
TLDR: A new method called Approximately Orthogonal Fine-Tuning (AOFT) improves the adaptation of pre-trained Vision Transformers (ViTs) for new tasks. It achieves this by...