ARG Logo

Abdelfattah Research Group

Home · Publications · Blog · Code CSL Logo Cornell Logo

Blog


Blogging is a new thing we're trying in our research group to share research ideas and results that are either not a good fit, or too preliminary for traditional peer-reviewed publication. We will often also share opinions, retrospectives, and other random content that could benefit from dissemination and feedback. All our blog posts have a comments section to enable interaction and perhaps even start new collaborations!

2026

Hierarchical SMC-SD: Composing Speculative Decoding Approaches

Yahya Emara, Mohamed Abdelfattah

2026

Code

Quantization for LLMs: A HW/SW Co-Design Perspective

Yuzong Chen

2026

Sequential Monte Carlo Speculative Decoding

Yahya Emara

2026

Paper · Code

Hardware Efficient Randomized SVD

Sean Mo, Chi-Chih Chang, Mohamed Abdelfattah

2026

Code

Rethinking Prefix Caching for Hybrid LLMs

Isabella Qiao, Crystal Zhou, Chi-Chih Chang, Mohamed Abdelfattah

2026

2025

KV-Cache Refresh Methods for Long Generation

Yahya Emara, Woojeong Kim, Mohamed Abdelfattah

2025

Paper · Code

Simulating large systems with Regression Language Models

Yash Akhauri, Xingyou Song

2025

Paper · Code

Extending MXFP4 and NVFP4 with Redundant Zero Remapping (RaZeR) for Accurate 4-bit LLM Quantization

Yuzong Chen, Xilai Dai, Mohamed Abdelfattah

2025

MXFP4 · NVFP4 · BitMoD · RaZeR

Quantifying GPU Performance overhead for Quantizated LLMs

Xilai Dai, Mohamed Abdelfattah

2025

AWQ · Marlin · Flute

Search

Tags

llm () hardware () fpga () software () dnn compression () gpu () automl () quantization ()