Logits are All We Need to Adapt Closed Models

Gaurush Hiranandani, Haolun Wu, Subhojyoti Mukherjee, Sanmi Koyejo
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:23261-23289, 2025.

Abstract

Many commercial Large Language Models (LLMs) are often closed-source, limiting developers to prompt tuning for aligning content generation with specific applications. While these models currently do not provide access to token logits, we argue that if such access were available, it would enable more powerful adaptation techniques beyond prompt engineering. In this paper, we propose a token-level probability reweighting framework that, given access to logits and a small amount of task-specific data, can effectively steer black-box LLMs toward application-specific content generation. Our approach views next-token prediction through the lens of supervised classification. We show that aligning black-box LLMs with task-specific data can be formulated as a label noise correction problem, leading to Plugin model – an autoregressive probability reweighting model that operates solely on logits. We provide theoretical justification for why reweighting logits alone is sufficient for task adaptation. Extensive experiments with multiple datasets, LLMs, and reweighting models demonstrate the effectiveness of our method, advocating for broader access to token logits in closed-source models. We provide our code at this https URL.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-hiranandani25a, title = {Logits are All We Need to Adapt Closed Models}, author = {Hiranandani, Gaurush and Wu, Haolun and Mukherjee, Subhojyoti and Koyejo, Sanmi}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {23261--23289}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/hiranandani25a/hiranandani25a.pdf}, url = {https://proceedings.mlr.press/v267/hiranandani25a.html}, abstract = {Many commercial Large Language Models (LLMs) are often closed-source, limiting developers to prompt tuning for aligning content generation with specific applications. While these models currently do not provide access to token logits, we argue that if such access were available, it would enable more powerful adaptation techniques beyond prompt engineering. In this paper, we propose a token-level probability reweighting framework that, given access to logits and a small amount of task-specific data, can effectively steer black-box LLMs toward application-specific content generation. Our approach views next-token prediction through the lens of supervised classification. We show that aligning black-box LLMs with task-specific data can be formulated as a label noise correction problem, leading to Plugin model – an autoregressive probability reweighting model that operates solely on logits. We provide theoretical justification for why reweighting logits alone is sufficient for task adaptation. Extensive experiments with multiple datasets, LLMs, and reweighting models demonstrate the effectiveness of our method, advocating for broader access to token logits in closed-source models. We provide our code at this https URL.} }
Endnote
%0 Conference Paper %T Logits are All We Need to Adapt Closed Models %A Gaurush Hiranandani %A Haolun Wu %A Subhojyoti Mukherjee %A Sanmi Koyejo %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-hiranandani25a %I PMLR %P 23261--23289 %U https://proceedings.mlr.press/v267/hiranandani25a.html %V 267 %X Many commercial Large Language Models (LLMs) are often closed-source, limiting developers to prompt tuning for aligning content generation with specific applications. While these models currently do not provide access to token logits, we argue that if such access were available, it would enable more powerful adaptation techniques beyond prompt engineering. In this paper, we propose a token-level probability reweighting framework that, given access to logits and a small amount of task-specific data, can effectively steer black-box LLMs toward application-specific content generation. Our approach views next-token prediction through the lens of supervised classification. We show that aligning black-box LLMs with task-specific data can be formulated as a label noise correction problem, leading to Plugin model – an autoregressive probability reweighting model that operates solely on logits. We provide theoretical justification for why reweighting logits alone is sufficient for task adaptation. Extensive experiments with multiple datasets, LLMs, and reweighting models demonstrate the effectiveness of our method, advocating for broader access to token logits in closed-source models. We provide our code at this https URL.
APA
Hiranandani, G., Wu, H., Mukherjee, S. & Koyejo, S.. (2025). Logits are All We Need to Adapt Closed Models. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:23261-23289 Available from https://proceedings.mlr.press/v267/hiranandani25a.html.

Related Material