AI Paper Atlas

Research map / Architectures and efficiency / Quantization

Tequila: Deadzone-free Ternary Quantization for Large Language Models

Hong Huang, Decheng Wu, Rui Cen, Guanghua Yu, Zonghang Li, Kai Liu, Jianchen Zhu, Peng Chen, Xue Liu, Dapeng Wu

ICLR 2026 · Architectures and efficiency › Quantization · approach: Post-training quantization of LLMs

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
reactivate dead weights in place as signed minima (Minima Reactivation) (ablation)AbsMean ternary, LLaMA-3.2-1B · ARC-eaccuracy: +1.2%Improved
replace STE-based minima with differentiable reactivation (lambda*w) (ablation)Minima Reactivation · ARC-eaccuracy: +1.1%Improved
add mixed gradients (hybrid weight and bias role) (ablation)Tequila w/o Mixed Gradients · ARC-eaccuracy: +1.9%Improved
add Tequila reactivation to AbsMeanAbsMean ternary QAT, LLaMA-3.2-1B · Average of 6 benchmarksaccuracy: 0.445 -> 0.471Improved
add Tequila reactivation to AbsMeanAbsMean ternary QAT, LLaMA-3.2-3B · Average of 6 benchmarksaccuracy: 0.510 -> 0.530Improved
add Tequila reactivation to AbsMeanAbsMean ternary QAT, LLaMA-3.2-3B · GPQAaccuracy: 0.301 -> 0.303No clear effect
add Tequila reactivationAbsMean ternary QAT, Qwen3-4B · Average of 5 benchmarksaccuracy: 0.5647 -> 0.5985Improved
use learnable-scale quantizers (LSQ, SEQ, DLT) instead of static AbsMean (ablation)ternary QAT, LLaMA-3.2-1B · Average of 6 benchmarksaccuracy: 0.445 -> 0.354-0.370Hurt
vary reactivation parameter lambda (ablation)Tequila · Average accuracyaccuracyNo clear effect
change quantization granularity from per-tensor to per-channel (ablation)Tequila, LLaMA-3.2-1B · Average accuracyaccuracy: 0.463 -> 0.471Improved
train Tequila on 10B tokens vs ternary LLMs trained on 100B tokensSpectra-3.9B · Average of 5 benchmarks (3B)accuracy: 0.567 -> 0.576Improved
add dead-weight bias term at inferenceTequila · inference speed (CPU)speedup vs BF16: 3.0x speedup, bias overhead < 0.1%No clear effect

Later papers that compared with this method

1 later accepted paper reports Tequila as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy