Is EAQuant superseded?

EAQuant (Mixture-of-experts routing): superseded — cited as a baseline and beaten by newer methods. 1 paper(s) critique it, 1 beat it on benchmarks — #122 of 1370 most-superseded. Sub-problem: cluster led by MoEQuant. Newer alternatives in the same sub-problem include BitsMoE, GEMQ, KBVQ-MoE, MC# (Mixture-Compressor-sharp).

Mixture-of-experts routing

EAQuant

EAQuant: Enhancing Post-Training Quantization for MoE Models via Expert-Aware Optimization

Superseded baseline#122 of 1,370 most-superseded · first seen Jun 16, 2025

Superseded — cited as a baseline and beaten by newer methods

1 papers critique it · 1 beat it on benchmarks

What papers say

Verbatim critique sentences, each from a paper that cites EAQuant as a baseline.

EAQuant~fu2025eaquant aligns router logits and expert-selection probabilities before and after quantization, but its full-dimensional alignment may allocate optimization effort to low-ranked experts that rarely affect top-$k$ decisions.

— Value-and-Structure Alignment for Routing-Consistent Quantization of Mixture-of-Experts Models

Beaten on benchmarks

Head-to-head results where a newer method reports beating EAQuant. Values are copied from the source paper's tables — verify against the cited paper.

GEMQ beats EAQuant
76.10 vs 71.23
0-shot accuracy · [3-4 bits vs 3-16 bits (weight-activation)]
GEMQ: Global Expert-Level Mixed-Precision Quantization for MoE LLMs

What to use instead

Recent methods in the same sub-problem, not yet superseded in the knowledge base — arXiv benchmark leaders, not vetted production recommendations.