LLM quantization

DiffQ

Superseded baseline#71 of 80 most-superseded

Cited as a baseline — critiqued by newer work, not yet beaten on a benchmark here

1 papers critique it · 0 beat it on benchmarks

What papers say

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

found that STE can cause weight oscillations during training
MEC-Quant: Maximum Entropy Coding for Extremely Low Bit Quantization-Aware Training

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Recent methods in the same sub-problem, not yet superseded in the knowledge base — arXiv benchmark leaders, not vetted production recommendations.