IST Austria Distributed Algorithms and Systems Lab

university

https://ist.ac.at/en/research/alistarh-group/

https://github.com/IST-DASLab

AI & ML interests

None defined yet.

Recent Activity

BlackSamorez authored a paper 3 days ago

Quartet II: Accurate LLM Pre-Training in NVFP4 by Improved Unbiased Gradient Estimation

soroushtabesh authored a paper 3 days ago

Quartet II: Accurate LLM Pre-Training in NVFP4 by Improved Unbiased Gradient Estimation

BlackSamorez submitted a paper 3 days ago

Quartet II: Accurate LLM Pre-Training in NVFP4 by Improved Unbiased Gradient Estimation

View all activity

Papers

Quartet II: Accurate LLM Pre-Training in NVFP4 by Improved Unbiased Gradient Estimation

WUSH: Near-Optimal Adaptive Transforms for LLM Quantization

View all Papers

ISTA-DASLab 's Papers 9

Submitted by

Andrei Panferov

Quartet II: Accurate LLM Pre-Training in NVFP4 by Improved Unbiased Gradient Estimation

ISTA-DASLab

IST Austria Distributed Algorithms and Systems Lab

Submitted by

Jiale Chen

WUSH: Near-Optimal Adaptive Transforms for LLM Quantization

ISTA-DASLab

IST Austria Distributed Algorithms and Systems Lab

CAGE: Curvature-Aware Gradient Estimation For Accurate Quantization-Aware Training

ISTA-DASLab

IST Austria Distributed Algorithms and Systems Lab

Submitted by

Denis Kuznedelev

Bridging the Gap Between Promise and Performance for Microscaling FP4 Quantization

ISTA-DASLab

IST Austria Distributed Algorithms and Systems Lab

Submitted by

Jiale Chen

The Geometry of LLM Quantization: GPTQ as Babai's Nearest Plane Algorithm

ISTA-DASLab

IST Austria Distributed Algorithms and Systems Lab

3

Submitted by

Andrei Panferov

Quartet: Native FP4 Training Can Be Optimal for Large Language Models

ISTA-DASLab

IST Austria Distributed Algorithms and Systems Lab

Submitted by

Alistarh

QuEST: Stable Training of LLMs with 1-Bit Weights and Activations

ISTA-DASLab

IST Austria Distributed Algorithms and Systems Lab

Scalable Mechanistic Neural Networks

ISTA-DASLab

IST Austria Distributed Algorithms and Systems Lab

MARLIN: Mixed-Precision Auto-Regressive Parallel Inference on Large Language Models

ISTA-DASLab

IST Austria Distributed Algorithms and Systems Lab