QuIP (LLM quantization): superseded — cited as a baseline and beaten by newer methods. 3 paper(s) critique it, 4 beat it on benchmarks — #9 of 80 most-superseded. Sub-problem: cluster led by GPTQ. Newer alternatives in the same sub-problem include QVGGT, LFQ, ADMM-Q, OSAQ, SEPTQ.

Superseded baseline#9 of 80 most-superseded

QuIP

QuIP: 2-Bit Quantization of Large Language Models With Guarantees

LLM quantization · first seen Jul 25, 2023

superseded — cited as a baseline and beaten by newer methods

3 papers critique it · 4 beat it on benchmarks

What papers say

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

“It is a powerful method and achieves 2-bit level quantization, but its computation cost is a little expensive.”
— SEPTQ: A Simple and Effective Post-Training Quantization Paradigm for Large Language Models
“Table tab:ppl2048 also shows the importance of incoherence processing. without fine-tuning or lattice codebooks significantly outperforms OmniQuant and AWQ, which both rely on heuristics to reduce model outliers during quantization.”
— QuIP#: Even Better LLM Quantization with Hadamard Incoherence and Lattice Codebooks
“However, these predetermined rotations cannot adapt to specific models.”
— ButterflyQuant: Ultra-low-bit LLM Quantization through Learnable Orthogonal Butterfly Transforms

Beaten on benchmarks

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

SEPTQ beats QuIP · perplexity [2-bit]
53.75 vs 177.40
SEPTQ: A Simple and Effective Post-Training Quantization Paradigm for Large Language Models
SEPTQ beats QuIP · perplexity [3-bit]
30.37 vs 30.92
SEPTQ: A Simple and Effective Post-Training Quantization Paradigm for Large Language Models
SEPTQ beats QuIP · perplexity [4-bit]
16.37 vs 16.38
SEPTQ: A Simple and Effective Post-Training Quantization Paradigm for Large Language Models
SEPTQ beats QuIP · Accuracy [2-bit, 7B]
48.91 vs 44.50
SEPTQ: A Simple and Effective Post-Training Quantization Paradigm for Large Language Models
SEPTQ beats QuIP · Accuracy [2-bit, 13B]
59.62 vs 51.57
SEPTQ: A Simple and Effective Post-Training Quantization Paradigm for Large Language Models
SEPTQ beats QuIP · Accuracy [2-bit, 30B]
63.07 vs 54.78
SEPTQ: A Simple and Effective Post-Training Quantization Paradigm for Large Language Models
SEPTQ beats QuIP · Accuracy [3-bit, 7B]
60.50 vs 58.73
SEPTQ: A Simple and Effective Post-Training Quantization Paradigm for Large Language Models
SEPTQ beats QuIP · Accuracy [3-bit, 13B]
64.21 vs 63.55
SEPTQ: A Simple and Effective Post-Training Quantization Paradigm for Large Language Models
SEPTQ beats QuIP · Accuracy [3-bit, 30B]
66.77 vs 64.38
SEPTQ: A Simple and Effective Post-Training Quantization Paradigm for Large Language Models
FrameQuant beats QuIP · Top-1 accuracy [2-bit quantization, ViT-T]
8.92 vs 1.42
FrameQuant: Flexible Low-Bit Quantization for Transformers
FrameQuant beats QuIP · Top-1 accuracy [2-bit quantization, ViT-S]
48.10 vs 21.98
FrameQuant: Flexible Low-Bit Quantization for Transformers
FrameQuant beats QuIP · Top-1 accuracy [2-bit quantization, ViT-S/32]
41.16 vs 19.00
FrameQuant: Flexible Low-Bit Quantization for Transformers

Newer alternatives

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