AI Paper Atlas

Research map / Architectures and efficiency / Quantization

CAT-Q: Cost-efficient and Accurate Ternary Quantization for LLMs

Shigeng Wang, Chao Li, Yangyuxuan Kang, Jiawei Fan, Anbang Yao

ICML 2026 · Architectures and efficiency › Quantization · approach: KV cache quantization · 0 citations (OpenAlex)

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
add learnable modulation (three factors) over static approximation (ablation)CAT-Q, Qwen3-4B · Commonsense zero-shotaverage accuracy (%): 40.16 -> 57.06Improved
add softened ternarization on top of learnable modulation (ablation)CAT-Q, Qwen3-4B · Commonsense zero-shotaverage accuracy (%): 53.19 -> 57.06Improved
add softened ternarization on top of learnable modulation (ablation)CAT-Q, Llama2-7B · Commonsense zero-shotaverage accuracy (%): 55.97 -> 57.69Improved
learnable modulation instead of direct learning of alpha and threshold (ablation)CAT-Q, Qwen3-4B · Commonsense zero-shotaverage accuracy (%): 46.94 -> 57.06Improved
hard ternarization stage after differentiable stage (gamma 0.8 vs 1.0) (ablation)CAT-Q, Qwen3-4B · Commonsense zero-shotaverage accuracy (%): 55.02 -> 57.06Improved
set s0 to 30 instead of 1000 (ablation)CAT-Q, Qwen3-4B · Commonsense zero-shotaverage accuracy (%): 50.82 -> 57.06Improved
drop mu from weight reconstruction (W ~ alpha T) (ablation)CAT-Q, Qwen3-4B · Commonsense zero-shotaverage accuracy (%): 56.59 -> 57.06Improved
learnable threshold instead of hand-crafted 0.5 (ablation)CAT-Q, Qwen3-4B · Commonsense zero-shotaverage accuracy (%): 55.53 -> 57.06Improved
increase calibration samples from 512 to 1024 (ablation)CAT-Q, Qwen3-4B · Commonsense zero-shotaverage accuracy (%): 57.06 -> 58.80Improved
increase calibration samples from 512 to 1024 (ablation)CAT-Q, Llama2-7B · Commonsense zero-shotaverage accuracy (%): 57.69 -> 60.50Improved
use group size 128 instead of channel-wise (ablation)CAT-Q, Qwen3-4B · Commonsense zero-shotaverage accuracy (%): 55.06 -> 57.06Improved
use CAT-Q ternary format instead of Q4_K_M in llama.cppQwen3-4B · GPU decoding throughputtokens/s: 225.91 -> 320.34Improved

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