SoftSRV: Learn to Generate Targeted Synthetic Data
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915137874558976 |
|---|---|
| author | DeSalvo, Giulia Kagy, Jean-Fracois Karydas, Lazaros Rostamizadeh, Afshin Kumar, Sanjiv |
| author_facet | DeSalvo, Giulia Kagy, Jean-Fracois Karydas, Lazaros Rostamizadeh, Afshin Kumar, Sanjiv |
| contents | We present a novel framework, SoftSRV, that is used to generate targeted synthetic fine-tuning data for improving task-specific model performance. Given a sample from a target distribution, our proposed framework uses a data-driven loss minimization approach to steer a frozen large language model (LLM) to generate synthetic sequences that are similar to those from the target distribution. SoftSRV provides a practical improvement over common prompt engineering approaches that rely on human-engineered prompt-templates, which can be idiosyncratic, labor-intensive to craft, and may need to be specialized per domain. We empirically evaluate our method against standard baselines guiding a large LLM to generate synthetic data to fine-tune a smaller language model on three different domains (coding, math, reasoning). We perform these evaluations without any particular specialization of the framework to each domain, emphasizing the generality of our approach. We find that SoftSRV improves upon typical prompt engineering approaches, generating targeted data that leads to fine-tuned models with significantly better task-specific performance. In addition, SoftSRV-generated data better matches the target distribution according to the MAUVE similarity metric. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_16534 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | SoftSRV: Learn to Generate Targeted Synthetic Data DeSalvo, Giulia Kagy, Jean-Fracois Karydas, Lazaros Rostamizadeh, Afshin Kumar, Sanjiv Machine Learning We present a novel framework, SoftSRV, that is used to generate targeted synthetic fine-tuning data for improving task-specific model performance. Given a sample from a target distribution, our proposed framework uses a data-driven loss minimization approach to steer a frozen large language model (LLM) to generate synthetic sequences that are similar to those from the target distribution. SoftSRV provides a practical improvement over common prompt engineering approaches that rely on human-engineered prompt-templates, which can be idiosyncratic, labor-intensive to craft, and may need to be specialized per domain. We empirically evaluate our method against standard baselines guiding a large LLM to generate synthetic data to fine-tune a smaller language model on three different domains (coding, math, reasoning). We perform these evaluations without any particular specialization of the framework to each domain, emphasizing the generality of our approach. We find that SoftSRV improves upon typical prompt engineering approaches, generating targeted data that leads to fine-tuned models with significantly better task-specific performance. In addition, SoftSRV-generated data better matches the target distribution according to the MAUVE similarity metric. |
| title | SoftSRV: Learn to Generate Targeted Synthetic Data |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2410.16534 |