About
Research & Development
Nested-ReFT: Boosting LLM Fine-Tuning Efficiency with Layer Skipping
TLDR: Nested-ReFT is a novel framework designed to make Reinforcement Learning for Large Language Model (LLM) fine-tuning more computationally efficient. It achieves this by...
Research & Development
Enhancing Time Series Classification with Neural Collapse-Driven Semi-Supervised Learning
TLDR: rETF-semiSL is a new semi-supervised learning framework for time series data that pre-trains deep neural networks to enforce Neural Collapse, a state where...
Research & Development
PakBBQ: Uncovering and Addressing Language Model Biases in Pakistan’s Diverse Cultural Landscape
TLDR: PakBBQ is a new, culturally adapted bias benchmark for Large Language Models (LLMs) in Pakistan. It features over 17,000 English and Urdu question-answer...
Research & Development
Multi-Region Fusion Decoding: Enhancing Factual Accuracy in AI Image Descriptions
TLDR: MRFD (Multi-Region Fusion Decoding) is a novel, training-free method designed to reduce 'hallucinations' in Large Vision-Language Models (LVLMs). It works by identifying salient...
Research & Development
eMamba: Accelerating Mamba Models for Efficient Edge Computing
TLDR: eMamba is a new hardware acceleration framework designed to efficiently deploy Mamba models on resource-constrained edge devices. It achieves this by replacing complex...
Research & Development
Boosting Semantic Segmentation Performance with Smart Label Deformations
TLDR: NSegment+ is a novel data augmentation framework for semantic segmentation that improves model robustness against subtle 'implicit' label noise. It achieves this by...