Rethinking Output Alignment For 1-bit Post-Training Quantization of Large Language Models
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866914568376156160 |
|---|---|
| author | Hoang, Dung Anh Pham, Cuong Nguyen, Cuong le, Trung Cai, Jianfei Do, Thanh-Toan |
| author_facet | Hoang, Dung Anh Pham, Cuong Nguyen, Cuong le, Trung Cai, Jianfei Do, Thanh-Toan |
| contents | Large Language Models (LLMs) deliver strong performance across a wide range of NLP tasks, but their massive sizes hinder deployment on resource-constrained devices. To reduce their computational and memory burden, various compression techniques have been proposed, including quantization, pruning, and knowledge distillation. Among these, post-training quantization (PTQ) is widely adopted for its efficiency, as it requires no retraining and only a small dataset for calibration, enabling low-cost deployment. Recent advances for post-training quantization have demonstrated that even near 4-bit methods can maintain most of the original model performance. However, 1-bit quantization remains particularly challenging. A common strategy in 1-bit quantization is to determine binary weights by matching full-precision parameters, following a weight-driven criterion. However, this objective is not directly aligned with the quantized model's objective, which is to preserve the model's output behavior under the impact of quantization. A natural alternative is to adopt output-driven criteria that minimize discrepancies in model outputs using calibration data. Surprisingly, naive output-driven approaches often perform even worse in the 1-bit regime.
In this paper, we show that this failure arises from two fundamental issues: error accumulation across layers and, more critically, \emph{anisotropic distortion} of the representation space. Based on these insights, we propose a novel PTQ method for 1-bit LLMs that explicitly addresses these issues while maintaining computational efficiency. Extensive experiments demonstrate that our approach consistently outperforms existing 1-bit PTQ methods. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_21651 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Rethinking Output Alignment For 1-bit Post-Training Quantization of Large Language Models Hoang, Dung Anh Pham, Cuong Nguyen, Cuong le, Trung Cai, Jianfei Do, Thanh-Toan Machine Learning Large Language Models (LLMs) deliver strong performance across a wide range of NLP tasks, but their massive sizes hinder deployment on resource-constrained devices. To reduce their computational and memory burden, various compression techniques have been proposed, including quantization, pruning, and knowledge distillation. Among these, post-training quantization (PTQ) is widely adopted for its efficiency, as it requires no retraining and only a small dataset for calibration, enabling low-cost deployment. Recent advances for post-training quantization have demonstrated that even near 4-bit methods can maintain most of the original model performance. However, 1-bit quantization remains particularly challenging. A common strategy in 1-bit quantization is to determine binary weights by matching full-precision parameters, following a weight-driven criterion. However, this objective is not directly aligned with the quantized model's objective, which is to preserve the model's output behavior under the impact of quantization. A natural alternative is to adopt output-driven criteria that minimize discrepancies in model outputs using calibration data. Surprisingly, naive output-driven approaches often perform even worse in the 1-bit regime. In this paper, we show that this failure arises from two fundamental issues: error accumulation across layers and, more critically, \emph{anisotropic distortion} of the representation space. Based on these insights, we propose a novel PTQ method for 1-bit LLMs that explicitly addresses these issues while maintaining computational efficiency. Extensive experiments demonstrate that our approach consistently outperforms existing 1-bit PTQ methods. |
| title | Rethinking Output Alignment For 1-bit Post-Training Quantization of Large Language Models |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2512.21651 |