Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Fitting Is Not Enough: Smoothness in Extremely Quantized LLMs

About

Large language models (LLMs) achieve strong performance but incur high deployment costs, motivating extremely low-bit but lossy quantization. Existing quantization algorithms mainly focus on improving the numerical accuracy of forward computation to eliminate performance degradation. In this paper, we show that extremely quantized LLMs suffer from systematic smoothness degradation beyond numerical precision loss. Through a smoothness proxy, we observe that such degradation becomes increasingly severe as the quantization bit-width decreases. Furthermore, based on sequence neighborhood modeling, we find that quantized models exhibit a rapid reduction of effective token candidates within the prediction neighborhood, which directly leads to a sparser decoding tree and degraded generation quality. To validate it, we introduce a simple smoothness-preserving principle in both post-training quantization and quantization-aware training, and demonstrate that preserving smoothness brings additional gains beyond numerical accuracy. The core goal of this paper is to highlight smoothness preservation as an important design consideration for future extreme quantization methods. Code is available at https://github.com/xuyuzhuang11/FINE.

Yuzhuang Xu, Xu Han, Yuxuan Li, Pengzhan Li, Wanxiang Che• 2026

Related benchmarks

TaskDatasetResultRank
Language ModelingWikiText-2--
2320
Language ModelingC4
Perplexity10.87
1688
Question AnsweringARC Challenge
Accuracy (ARC)32.59
598
Question AnsweringPIQA
Accuracy71.71
505
Commonsense ReasoningWinoGrande
Accuracy58.17
453
Mathematical ReasoningMathQA--
354
Language ModelingWiki2
PPL9.25
326
Question AnsweringOpenBookQA
Accuracy34.2
305
Reading ComprehensionBoolQ
Accuracy (BoolQ)62.05
228
Commonsense ReasoningPIQA
Accuracy53.54
213
Showing 10 of 17 rows

Other info

Follow for update