Latticethe structure behind the noise

Papers Digest Topics Selected Labs Collections FAQ

Created by Flynn Lachendro

Papers Digest Topics Labs Saved

Search

Search papers, labs, and topics across Lattice.

Built by Flynn Lachendro·𝕏 / Twitter·RSS··FAQ·Glossary·Privacy

Saurav Jha | Lattice

Saurav Jha

Papers on Lattice

1

Total citations

0

Topics

2

h-index

1

Publication activitypapers/week, last 8 weeks

Research focus

Architecture Design (Transformers, SSMs, MoE) (1)Inference & Quantization (1)

Frequent co-authors

M. Hashemzadeh (1)Maryam Hashemzadeh (1)Ali Saheb Pasand (1)Alipanah Parviz (1)

Papers (1)

Apr 6, 2026

MilaApr 6, 2026·also AI Center, B-Instruct-2507. Here, McGill

REAM: Merging Improves Pruning of Experts in LLMs

Merging experts in MoE LLMs can actually *improve* performance compared to pruning, offering a new path to compression that preserves capabilities.

Saurav Jha, M. Hashemzadeh, Maryam Hashemzadeh +5

Architecture Design (Transformers, SSMs, MoE)Inference & Quantization

Min-Joong Lee (1)

Boris Knyazev (1)