Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2601.10416 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917205595127808 |
|---|---|
| author | Shen, Tiesunlong Mao, Rui Wang, Jin Sun, Heming Zhang, Jian Zhang, Xuejie Cambria, Erik |
| author_facet | Shen, Tiesunlong Mao, Rui Wang, Jin Sun, Heming Zhang, Jian Zhang, Xuejie Cambria, Erik |
| contents | Aligning Large Language Models (LLMs) with human preferences is critical, yet traditional fine-tuning methods are computationally expensive and inflexible. While test-time alignment offers a promising alternative, existing approaches often rely on distorted trajectory-level signals or inefficient sampling, fundamentally capping performance and failing to preserve the generative diversity of the base model. This paper introduces LLMdoctor, a novel framework for efficient test-time alignment that operates via a patient-doctor paradigm. It integrates token-level reward acquisition with token-level flow-guided preference optimization (TFPO) to steer a large, frozen patient LLM with a smaller, specialized doctor model. Unlike conventional methods that rely on trajectory-level rewards, LLMdoctor first extracts fine-grained, token-level preference signals from the patient model's behavioral variations. These signals then guide the training of the doctor model via TFPO, which establishes flow consistency across all subtrajectories, enabling precise token-by-token alignment while inherently preserving generation diversity. Extensive experiments demonstrate that LLMdoctor significantly outperforms existing test-time alignment methods and even surpasses the performance of full fine-tuning approaches like DPO. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_10416 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | LLMdoctor: Token-Level Flow-Guided Preference Optimization for Efficient Test-Time Alignment of Large Language Models Shen, Tiesunlong Mao, Rui Wang, Jin Sun, Heming Zhang, Jian Zhang, Xuejie Cambria, Erik Artificial Intelligence Aligning Large Language Models (LLMs) with human preferences is critical, yet traditional fine-tuning methods are computationally expensive and inflexible. While test-time alignment offers a promising alternative, existing approaches often rely on distorted trajectory-level signals or inefficient sampling, fundamentally capping performance and failing to preserve the generative diversity of the base model. This paper introduces LLMdoctor, a novel framework for efficient test-time alignment that operates via a patient-doctor paradigm. It integrates token-level reward acquisition with token-level flow-guided preference optimization (TFPO) to steer a large, frozen patient LLM with a smaller, specialized doctor model. Unlike conventional methods that rely on trajectory-level rewards, LLMdoctor first extracts fine-grained, token-level preference signals from the patient model's behavioral variations. These signals then guide the training of the doctor model via TFPO, which establishes flow consistency across all subtrajectories, enabling precise token-by-token alignment while inherently preserving generation diversity. Extensive experiments demonstrate that LLMdoctor significantly outperforms existing test-time alignment methods and even surpasses the performance of full fine-tuning approaches like DPO. |
| title | LLMdoctor: Token-Level Flow-Guided Preference Optimization for Efficient Test-Time Alignment of Large Language Models |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2601.10416 |