QTIP (LLM quantization): superseded — cited as a baseline and beaten by newer methods. 1 paper(s) critique it, 2 beat it on benchmarks — #25 of 80 most-superseded. Sub-problem: cluster led by RTN. Newer alternatives in the same sub-problem include STaR-Quant, Timestep-Aware SVDQuant-GPTQ, BWLA, Bit-by-Bit, Benford-Quant.

Superseded baseline#25 of 80 most-superseded

QTIP

QTIP: Quantization with Trellises and Incoherence Processing

LLM quantization · first seen Jun 17, 2024

superseded — cited as a baseline and beaten by newer methods

1 papers critique it · 2 beat it on benchmarks

What papers say

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

“Although QTIP shows significant improvement over TCQ, it still suffers from high computational complexity.”
— CCQ: Convolutional Code for Extreme Low-bit Quantization in LLMs

Beaten on benchmarks

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

GLVQ-8D beats QTIP · ARC-Challenge [2-bit quantization, Llama 2-13B]
40.0 vs 39.2
Learning Grouped Lattice Vector Quantizers for Low-Bit LLM Compression
GLVQ-32D beats QTIP · Perplexity [2-bit, Llama 2-7B]
5.41 vs 5.91
Learning Grouped Lattice Vector Quantizers for Low-Bit LLM Compression
GLVQ-32D beats QTIP · Perplexity [2-bit, Llama 2-70B]
3.36 vs 3.78
Learning Grouped Lattice Vector Quantizers for Low-Bit LLM Compression
ICQuant^SK-5% beats QTIP · C4 [Llama2-7B, ctx. 4096, 4.3 bits]
6.70 vs 6.71
ICQuant: Index Coding enables Low-bit LLM Quantization
ICQuant^SK-5% beats QTIP · Wiki2 [Llama2-13B, ctx. 4096, 4.3 bits]
4.61 vs 4.62
ICQuant: Index Coding enables Low-bit LLM Quantization
ICQuant^SK-5% beats QTIP · C4 [Llama2-13B, ctx. 4096, 4.3 bits]
6.09 vs 6.10
ICQuant: Index Coding enables Low-bit LLM Quantization
ICQuant^SK-5% beats QTIP · C4 [Llama2-13B, ctx. 4096, 3.3 bits]
6.26 vs 6.28
ICQuant: Index Coding enables Low-bit LLM Quantization
ICQuant^SK-8.25% beats QTIP · Wiki2 [Llama2-7B, ctx. 4096, 2.4 bits]
6.35 vs 6.82
ICQuant: Index Coding enables Low-bit LLM Quantization
ICQuant^SK-8.25% beats QTIP · C4 [Llama2-7B, ctx. 4096, 2.4 bits]
8.25 vs 8.96
ICQuant: Index Coding enables Low-bit LLM Quantization
ICQuant^SK-8.25% beats QTIP · C4 [Llama2-13B, ctx. 4096, 2.4 bits]
7.25 vs 7.39
ICQuant: Index Coding enables Low-bit LLM Quantization
ICQuant^SK-8.25% beats QTIP · Wiki2 [Llama2-70B, ctx. 4096, 2.4 bits]
3.86 vs 3.87
ICQuant: Index Coding enables Low-bit LLM Quantization
ICQuant^SK-8.25% beats QTIP · C4 [Llama2-70B, ctx. 4096, 2.4 bits]
5.61 vs 5.70
ICQuant: Index Coding enables Low-bit LLM Quantization

Newer alternatives

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