efficient-ai

Long-RL: Scaling RL to Long Sequences (NeurIPS 2025)

EfficientSAM3 compresses SAM3 into lightweight, edge-friendly models via progressive knowledge distillation for fast promptable concept segmentation and tracking.

[TMLR 2026] Survey: https://arxiv.org/pdf/2507.20198

[ICML 2026] GRACE: deployable INT4 VLM via quantization-aware distillation.

A High-Efficiency System of Large Language Model Based Search Agents

Official Repo for SparseLLM: Global Pruning of LLMs (NeurIPS 2024)

Official PyTorch implementation of the paper "Dataset Distillation via the Wasserstein Metric" (ICCV 2025).

TinyML and Efficient Deep Learning Computing | MIT 6.S965/6.5940

CVPR 2026 (Main) | MambaLiteUNet: Cross-Gated Adaptive Feature Fusion for Robust Skin Lesion Segmentation

Code for paper "Joint Architecture Design and Workload Partitioning for DNN Inference on Industrial IoT Clusters"

This is a repository accompanying the survey Edge AI Meets LLM (coming soon), containing a comprehensive list of papers, codebases, toolchains, and open-source frameworks. It is intended to serve as a handbook for researchers and developers interested in Edge/Mobile LLMs.

Dynamic Attention Mask (DAM) generate adaptive sparse attention masks per layer and head for Transformer models, enabling long-context inference with lower compute and memory overhead without fine-tuning.

Code for paper "EdgeKE: An On-Demand Deep Learning IoT System for Cognitive Big Data on Industrial Edge Devices"

Official PyTorch implementation of the paper "Towards Adversarially Robust Dataset Distillation by Curvature Regularization" (AAAI 2025).

Curated list of efficiency mechanisms for Video LLMs (companion to our survey)

A deep learning framework that implements Early Exit strategies in Convolutional Neural Networks (CNNs) using Deep Q-Learning (DQN). This project enhances computational efficiency by dynamically determining the optimal exit point in a neural network for image classification tasks on CIFAR-10.

Research-ready and production-friendly neural network pruning for PyTorch—transparent methods, reproducible baselines, and deployment metrics to compress models for real-world use.

⚡ Fast, concise, LLM-first Generative UI language

Transformer (GPT) implemented from scratch in C++. Runs on modest hardware with complete mathematical derivations and optimized tensor operations.