Likelihood Variance as Text Importance for Resampling Texts to Map Language Models
Fuente:
arXiv
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918028983140352 |
|---|---|
| author | Oyama, Momose Kishino, Ryo Yamagiwa, Hiroaki Shimodaira, Hidetoshi |
| author_facet | Oyama, Momose Kishino, Ryo Yamagiwa, Hiroaki Shimodaira, Hidetoshi |
| contents | We address the computational cost of constructing a model map, which embeds diverse language models into a common space for comparison via KL divergence. The map relies on log-likelihoods over a large text set, making the cost proportional to the number of texts. To reduce this cost, we propose a resampling method that selects important texts with weights proportional to the variance of log-likelihoods across models for each text. Our method significantly reduces the number of required texts while preserving the accuracy of KL divergence estimates. Experiments show that it achieves comparable performance to uniform sampling with about half as many texts, and also facilitates efficient incorporation of new models into an existing map. These results enable scalable and efficient construction of language model maps. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_15428 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Likelihood Variance as Text Importance for Resampling Texts to Map Language Models Oyama, Momose Kishino, Ryo Yamagiwa, Hiroaki Shimodaira, Hidetoshi Computation and Language We address the computational cost of constructing a model map, which embeds diverse language models into a common space for comparison via KL divergence. The map relies on log-likelihoods over a large text set, making the cost proportional to the number of texts. To reduce this cost, we propose a resampling method that selects important texts with weights proportional to the variance of log-likelihoods across models for each text. Our method significantly reduces the number of required texts while preserving the accuracy of KL divergence estimates. Experiments show that it achieves comparable performance to uniform sampling with about half as many texts, and also facilitates efficient incorporation of new models into an existing map. These results enable scalable and efficient construction of language model maps. |
| title | Likelihood Variance as Text Importance for Resampling Texts to Map Language Models |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2505.15428 |