Research map / Architectures and efficiency / Quantization
PT$^2$-LLM: Post-Training Ternarization for Large Language Models
Xianglong Yan, ChengZhu Bao, Zhiteng Li, Tianao Zhang, Kaicheng Yang, Haotong Qin, Ruobing Xie, Xingwu Sun, Yulun Zhang
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add Iterative Ternary Fitting (ablation) | PT2-LLM on LLaMA-2-7B · WikiText2 | perplexity: 22.88 -> 15.47 | Improved |
| add Activation-aware Grid Alignment (ablation) | PT2-LLM on LLaMA-2-7B · 7 zero-shot tasks | avg accuracy: 37.11 -> 42.86 | Improved |
| add ITF (ablation) | PT2-LLM on LLaMA-3-8B · WikiText2 | perplexity: 247.75 -> 83.76 | Improved |
| add AGA (ablation) | PT2-LLM on LLaMA-3-8B · WikiText2 | perplexity: 247.75 -> 47.83 | Improved |
| add AGA on top of ITF (ablation) | PT2-LLM on LLaMA-3-8B · WikiText2 | perplexity: 83.76 -> 32.19 | Improved |
| add SSR instead of no reordering (ablation) | PT2-LLM on LLaMA-2-7B · 7 zero-shot tasks | avg accuracy: 41.37 -> 43.33 | Improved |
| add SSR instead of no reordering (ablation) | PT2-LLM on LLaMA-3-8B · WikiText2 | perplexity: 112.83 -> 32.19 | Improved |
| use SSR instead of Hessian-based reordering (ablation) | PT2-LLM on LLaMA-2-7B · 7 zero-shot tasks | avg accuracy: 39.15 -> 43.33 | Improved |
| use random reordering instead of none (ablation) | PT2-LLM on LLaMA-2-7B · 7 zero-shot tasks | avg accuracy: 41.37 -> 40.86 | No clear effect |
| use Hessian-based reordering instead of none (ablation) | PT2-LLM on LLaMA-3-8B · WikiText2 | perplexity: 112.83 -> 35.86 | Improved |
| increase calibration set from 64 to 128 samples (ablation) | PT2-LLM on LLaMA-3-8B · WikiText2 | perplexity: 38.90 -> 32.19 | Improved |
| increase calibration set from 128 to 256 samples (ablation) | PT2-LLM on LLaMA-2-7B · 7 zero-shot tasks | avg accuracy: 43.33 -> 43.55 | No clear effect |
Later papers that compared with this method
1 later accepted paper reports PT2-LLM as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- TWLA: Achieving Ternary Weights and Low-Bit Activations for LLMs via Post-Training Quantization (ICML 2026 · better on 2 of 2 shared results)
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
- BiLLM: Pushing the Limit of Post-Training Quantization for LLMs (ICML 2024 · similarity 0.92)
- Exploring Post-training Quantization in LLMs from Comprehensive Study to Low Rank Compensation (AAAI 2024 · similarity 0.92)
- QLLM: Accurate and Efficient Low-Bitwidth Quantization for Large Language Models (ICLR 2024 · similarity 0.91)
- ResQ: Mixed-Precision Quantization of Large Language Models with Low-Rank Residuals (ICML 2025 · similarity 0.91)
- LeanQuant: Accurate and Scalable Large Language Model Quantization with Loss-error-aware Grid (ICLR 2025 · similarity 0.91)
- Evaluating Quantized Large Language Models (ICML 2024 · similarity 0.91)
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.
