Generative Active Testing: Efficient LLM Evaluation via Proxy Task Adaptation
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866915876176920576 |
|---|---|
| author | Ramakrishnan, Aashish Anantha Saeedi, Ardavan Hassanzadeh, Hamid Reza Mohaghegh, Fazlolah Lee, Dongwon |
| author_facet | Ramakrishnan, Aashish Anantha Saeedi, Ardavan Hassanzadeh, Hamid Reza Mohaghegh, Fazlolah Lee, Dongwon |
| contents | With the widespread adoption of pre-trained Large Language Models (LLM), there exists a high demand for task-specific test sets to benchmark their performance in domains such as healthcare and biomedicine. However, the cost of labeling test samples while developing new benchmarks poses a significant challenge, especially when expert annotators are required. Existing frameworks for active sample selection offer limited support for generative Question Answering tasks, where option dynamics can affect model decision boundaries. In this paper, we present Generative Active Testing (GAT), an uncertainty-aware acquisition framework leveraging LLMs as surrogates for informing the sample selection process. Using a novel Statement Adaptation Module, we modify generative tasks into a pseudo-classification format, enabling the capture of sample-level uncertainties across unlabeled candidates. Our zero-shot acquisition functions reduce estimation error by ~40% compared to traditional sampling baselines, offering a scalable solution for cost-effective model benchmarking. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_19264 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Generative Active Testing: Efficient LLM Evaluation via Proxy Task Adaptation Ramakrishnan, Aashish Anantha Saeedi, Ardavan Hassanzadeh, Hamid Reza Mohaghegh, Fazlolah Lee, Dongwon Computation and Language Artificial Intelligence I.2.7 With the widespread adoption of pre-trained Large Language Models (LLM), there exists a high demand for task-specific test sets to benchmark their performance in domains such as healthcare and biomedicine. However, the cost of labeling test samples while developing new benchmarks poses a significant challenge, especially when expert annotators are required. Existing frameworks for active sample selection offer limited support for generative Question Answering tasks, where option dynamics can affect model decision boundaries. In this paper, we present Generative Active Testing (GAT), an uncertainty-aware acquisition framework leveraging LLMs as surrogates for informing the sample selection process. Using a novel Statement Adaptation Module, we modify generative tasks into a pseudo-classification format, enabling the capture of sample-level uncertainties across unlabeled candidates. Our zero-shot acquisition functions reduce estimation error by ~40% compared to traditional sampling baselines, offering a scalable solution for cost-effective model benchmarking. |
| title | Generative Active Testing: Efficient LLM Evaluation via Proxy Task Adaptation |
| topic | Computation and Language Artificial Intelligence I.2.7 |
| url | https://arxiv.org/abs/2603.19264 |