Is DuQuant superseded?

DuQuant (LLM quantization): superseded — cited as a baseline and beaten by newer methods. 1 paper(s) critique it, 1 beat it on benchmarks — #39 of 80 most-superseded. Sub-problem: cluster led by GPTAQ. Newer alternatives in the same sub-problem include SplitQ, MARR, Quantization-aware Integrated Gradients (QIG), SPEED-Q, Quant-dLLM.

Method Drift›LLM quantization

Superseded baseline#39 of 80 most-superseded

DuQuant

DuQuant: Distributing Outliers via Dual Transformation Makes Stronger Quantized LLMs

LLM quantization · first seen Jun 3, 2024

superseded — cited as a baseline and beaten by newer methods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 DuQuant as a baseline.

“lin2024duquant manually design a zigzag permutation pattern in an attempt to distribute outlier weights throughout the network.”
— Exploring Model Invariance with Discrete Search for Ultra-Low-Bit Quantization

Beaten on benchmarks

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

SplitQ (Ours) beats DuQuant · Avg [W4A8 (7B)]
64.1 vs 53.4
Breaking Modality Heterogeneity in Low-Bit Quantization for Large Vision-Language Models
SplitQ (Ours) beats DuQuant · Avg [W4A4 (7B)]
62.9 vs 52.0
Breaking Modality Heterogeneity in Low-Bit Quantization for Large Vision-Language Models
SplitQ (Ours) beats DuQuant · Avg [W4A8 (13B)]
66.9 vs 65.0
Breaking Modality Heterogeneity in Low-Bit Quantization for Large Vision-Language Models
SplitQ (Ours) beats DuQuant · Avg [W4A4 (13B)]
66.4 vs 62.4
Breaking Modality Heterogeneity in Low-Bit Quantization for Large Vision-Language Models

Newer alternatives

Recent methods in the same sub-problem, not yet superseded in the knowledge base.