Ternary Quantization

Follow this topic Ask about this topic

Ternary quantization represents weights with three values, commonly −1, 0, and +1, to reduce model storage and computation. Methods include post-training quantization with learned scales, rotations, or GPTQ-style error compensation, as well as quantization-aware training, distillation, and sparse residual corrections; some also quantize activations. Deployment work develops compact packing and specialized kernels for CPUs, GPUs, FPGAs, and encrypted inference, since compressed weights do not automatically make execution faster. The central challenge is preserving capabilities across models and tasks while measuring real storage, accuracy, and runtime costs, including whether export and hardware implementations preserve the quantized model faithfully.

Latest papers

25 papers5 papers in the last four weeks, level with the four weeks before. 0.0% of all new papers.

Noise-shaped Sampling and Annealing for Scalable On-the-edge Training of Binary and Ternary AI Models
Oct 15, 2026Zihao Chen, Zhili Xiao, Rui-Jie Zhu +3Binary Neural Networks Ternary Quantization
Department of Electrical and Systems Engineering, Washington University in St. Louis, One Brookings Drive, St. Louis, 63130, MO, USA. · Department of Electrical and Computer Engineering, University of California, Santa Cruz, 1156 High Street, Santa Cruz, 95064, CA, USA. · Department of Computer Science and Engineering, Washington University in St. Louis, One Brookings Drive, St. Louis, 63130, MO, USA.
RoleQ: Role-Aligned Scales for Converting Language Models to Ternary
Oct 15, 2026Johannes Knittel, Alec Helbling, Hanspeter PfisterLLM Quantization Neural Network Quantization
Harvard University · Georgia Tech
SSR: Sparse Segment Reduction for Ternary GEMM Acceleration
Oct 6, 2026Adeline Pittet, Shien Zhu, Valérie Verdan +1Ternary Quantization LLM Inference Acceleration
Systems Group, ETH Zurich, Switzerland
Lost in the bf16 Cast: Exporting Ternary Language Models Can Revert Most Low-Learning-Rate Code Changes
Oct 6, 2026Avichal Sahai, Nishant Raj, Animesh SrivastavaLLM Quantization Ternary Quantization
Ofbusiness
Fiona: Accelerating FHE Inference with Packing-Aware Ternary Weights
Sep 28, 2026Yiteng Peng, Zhibo Liu, Dongwei Xiao +1Mixed-Precision Quantization Efficient Neural Network Inference
The Hong Kong University of Science and Technology · State Key Laboratory of Novel Software Technology, Nanjing University
Baseline Shape Decides the Verdict: A Controlled Re-Examination of Ternary Language Models at 60K Parameters
Sep 24, 2026Gautam VeldandaTernary Quantization Quantization-Aware Training
Independent Researcher
Breaking the 1.58-bit Barrier for Ternary LLMs
Sep 14, 2026Evangelos Georganas, Alexander Heinecke, Pradeep DubeyTernary Quantization LLM Compression
Intel Corporation
Scaling Post-Training Ternarisation to Qwen3-8B Capability Retention, Reproduction, Lossless Packing, and Packed Execution
Sep 8, 2026Anirudh Malik, M Sparsh Mehra, Poojith DevanLLM Quantization Ternary Quantization
OneBit AI
Post-Training Ternarization of Qwen3-4B Capability, Effective Bit Budget, Storage Compression, and Deployment
Sep 2, 2026Anirudh Malik, M Sparsh Mehra, Poojith DevanLLM Quantization Ternary Quantization
OneBit AI
QTEA: Ternary LLMs with Sparse Residual Salient Weight and By-Column Optimization
Aug 31, 2026Yipin Guo, Arun M George, Jie Fu +3LLM Quantization Ternary Quantization
University of Notre Dame
Capability-Stratified Degradation in Ternary Language Models
Aug 28, 2026Anirudh Malik, M Sparsh Mehra, Poojith DevanLLM Quantization Ternary Quantization
OneBit AI
Low-Rank Ternary Adaptation for Fine-Tuning Transformers
Aug 25, 2026Alexandru-Dragos Manolache, Yunqiang Li, Jan van GemertParameter-Efficient Fine-Tuning Low-Rank Adaptation
Delft University of Technology, The Netherlands · Amazon Development Center
Tied Trit-Planes: Constraining PTQTP to a Uniform Nine-Level Quantizer, with a Persistent Folded Format for Disk-Streamed Mixture-of-Experts Serving
Aug 9, 2026Matteo GrellaTernary Quantization Mixture-of-Experts Inference
Crisis24
Attend to Your Own Thoughts: Breaking the Barrier for Post-Training Quantization of Reasoning LLMs through the Lens of 1.58-Bit Quantization
Aug 2, 2026Shigeng Wang, Chao Li, Yangyuxuan Kang +2LLM Quantization Ternary Quantization
Intel Labs China
ExTernD: Expanded-Rank Ternary Decomposition Ternary LLM PTQ with Accuracy Approaching Any Quantization Level
Jul 15, 2026Chethan Reddy G. PLLM Quantization Ternary Quantization
ELiTeFormer: An Efficient Transformer for FPGAs
Jul 4, 2026Victor Agostinelli, Nicolas Bohm Agostini, Antonino TumeoLLM Quantization Ternary Quantization
Oregon State University Corvallis, Oregon, USA · Pacific Northwest National Laboratory Richland, Washington, USA
CAT-Q: Cost-efficient and Accurate Ternary Quantization for LLMs
Jun 25, 2026Shigeng Wang, Chao Li, Yangyuxuan Kang +2LLM Quantization Ternary Quantization
1Intel Labs China
Ternary Mamba: Grouped Quantization-Aware Training of W1.58A16 State Space Models
Jun 16, 2026Ramprasath Ganesaraja, Sahil Dilip Panse, Swathika NLLM Quantization Ternary Quantization
EdgeVerve Systems Limited
TWLA: Achieving Ternary Weights and Low-Bit Activations for LLMs via Post-Training Quantization
Jun 11, 2026Zhixiong Zhao, Zukang Xu, Zhixuan Chen +3Mixed-Precision Quantization LLM Quantization
Houmo AI, China · 2Southeast University, Nanjing, China
FTerViT: Fully Ternary Vision Transformer
May 20, 2026Szymon Ruciński, Pietro Bonazzi, Engin Türetken +3Ternary Quantization Vision Transformer
CSEM, Neuchâtel, Switzerland · ETH Zürich, Zurich, Switzerland
A Geometric Analysis of Sign-Magnitude Asymmetry in a ReLU + RMSNorm Block under Ternary Quantization
May 18, 2026Lei DongLLM Quantization Ternary Quantization
Independent Researcher
FairyFuse: Multiplication-Free LLM Inference on CPUs via Fused Ternary Kernels
Apr 22, 2026Fei Zuo, Xiaoyan Xi, Quanyi Zeng +2Efficient Neural Network Inference Ternary Quantization
BA TechWorks · Peking University
ITQ3_S: High-Fidelity 3-bit LLM Inference via Interleaved Ternary Quantization with Rotation-Domain Smoothing
Mar 30, 2026Edward J. YoonLLM Quantization Ternary Quantization
ButterflyMoE: Compression-Scalable Ternary Experts via Structured Butterfly Orbits
Jan 20, 2026Aryan KarmoreTernary Quantization Mixture-of-Experts Language Models
Indian Institute of Information Technology, Nagpur
TernaryLLM: Ternarized Large Language Model
Jun 11, 2024Tianqi Chen, Zhe Li, Weixiang Xu +6LLM Quantization Ternary Quantization
Institute of Automation, Chinese Academy of Sciences · School of Artificial Intelligence, University of Chinese Academy of Sciences · Advanced Micro Devices, Inc., Beijing, China +1

Oct 15, 2026J/K move · Enter open · S save

Noise-shaped Sampling and Annealing for Scalable On-the-edge Training of Binary and Ternary AI Models

Zihao Chen, Zhili Xiao, Rui-Jie Zhu, Yipeng Wang+2

Department of Electrical and Systems Engineering, Washington University in St. Louis, One Brookings Drive, St. Louis, 63130, MO, USA. · Department of Electrical and Computer Engineering, University of California, Santa Cruz, 1156 High Street, Santa Cruz, 95064, CA, USA. · Department of Computer Science and Engineering, Washington University in St. Louis, One Brookings Drive, St. Louis, 63130, MO, USA.

Binary Neural Networks Ternary Quantization On-Device Training LLM Quantization Quantization-Aware Training

Open paper AskPDF arXiv

The training loss of binary and ternary neural networks, expanded as a multilinear polynomial over the discrete weights, is structurally equivalent to the energy of a hierarchical high-order Ising model. We exploit this equivalence to apply Ising solvers to discrete-weight network training, while accounting for a fundamental difference in objectives: an Ising solver seeks the ground state of a fixed energy function, whereas training seeks solutions that generalize to unseen data. Our key contribution is

ΣΔ

-Sampler, an optimizer that reconciles these objectives by embedding a stochastic sampling-and-annealing step as a noisy quantizer inside a first-order

ΣΔ

noise-shaping loop. The loop pushes noise from weight quantization, high-order frustration, and mini-batch sampling out of the low-frequency adaptation band, so that discrete weight updates follow the smooth descent trajectories embedded in the rugged loss landscape. We validate the framework on a planted quadratic unconstrained binary optimization (QUBO) benchmark with a known minimizer; on CIFAR-10, where a

13

M-parameter ternary model comes within

1\%

of full-precision accuracy; on ImageNet-1k, where a ternary ResNet-18 comes within

0.4\%

Top-5 accuracy of a ternary baseline; on class-incremental continual learning, where prior-task accuracy is preserved; and on MatMul-free language models up to

2.7

B parameters, where it outperforms a straight-through-estimator baseline under matched continued pre-training. Compared with standard quantization-aware training,

ΣΔ

-Sampler reduces the training memory footprint by over

50\%

, delivers an order-of-magnitude speed-up, and reduces weight-update memory traffic by four orders of magnitude, making it attractive for on-device training and fine-tuning of large binary and ternary models.