GIFT: Guided Fine-Tuning and Transfer for Enhancing Instruction-Tuned Language Models
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866911640254939136 |
|---|---|
| author | Ruan, Zhiwen Du, Yichao Zheng, Jianjie Wang, Longyue Chen, Yun Li, Peng Su, Jinsong Liu, Yang Chen, Guanhua |
| author_facet | Ruan, Zhiwen Du, Yichao Zheng, Jianjie Wang, Longyue Chen, Yun Li, Peng Su, Jinsong Liu, Yang Chen, Guanhua |
| contents | A promising paradigm for adapting instruction-tuned language models is to learn task-specific updates on a pretrained base model and subsequently merge them into the instruction-tuned model. However, existing approaches typically treat the instruction-tuned model as a passive target that is only involved at the final merging stage, without guiding the training process. We propose GIFT (Guided Fine-Tuning and Transfer), a simple and efficient framework that incorporates guidance from the instruction model into task adaptation. GIFT fine-tunes a low-rank adapter on the pretrained base model using confidence signals derived from the instruction-tuned model. The learned adapter is then merged into the instruction-tuned model, yielding task-specialized models that preserve general instruction-following behavior. We evaluate GIFT on mathematical and knowledge-intensive benchmarks across multiple model families and scales. Results show that GIFT consistently outperforms direct fine-tuning and representative transfer-based baselines, while maintaining robust generalization and favorable test-time scaling behavior. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_01256 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | GIFT: Guided Fine-Tuning and Transfer for Enhancing Instruction-Tuned Language Models Ruan, Zhiwen Du, Yichao Zheng, Jianjie Wang, Longyue Chen, Yun Li, Peng Su, Jinsong Liu, Yang Chen, Guanhua Computation and Language A promising paradigm for adapting instruction-tuned language models is to learn task-specific updates on a pretrained base model and subsequently merge them into the instruction-tuned model. However, existing approaches typically treat the instruction-tuned model as a passive target that is only involved at the final merging stage, without guiding the training process. We propose GIFT (Guided Fine-Tuning and Transfer), a simple and efficient framework that incorporates guidance from the instruction model into task adaptation. GIFT fine-tunes a low-rank adapter on the pretrained base model using confidence signals derived from the instruction-tuned model. The learned adapter is then merged into the instruction-tuned model, yielding task-specialized models that preserve general instruction-following behavior. We evaluate GIFT on mathematical and knowledge-intensive benchmarks across multiple model families and scales. Results show that GIFT consistently outperforms direct fine-tuning and representative transfer-based baselines, while maintaining robust generalization and favorable test-time scaling behavior. |
| title | GIFT: Guided Fine-Tuning and Transfer for Enhancing Instruction-Tuned Language Models |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2605.01256 |