NExT-Guard: Streaming Safeguard without Token-Level Labels

Junfeng Fang, Nachuan Chen, Houcheng Jiang, Dan Zhang, Xiangnan He, Tat-Seng Chua, Xiang Wang
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:28984-29000, 2026.

Abstract

Large language models are increasingly deployed in streaming scenarios, rendering conventional post-hoc safeguards ineffective as they fail to interdict unsafe content in real-time. While streaming safeguards based on token-level supervised training could address this, they necessitate expensive annotations and suffer from severe overfitting. In this work, we challenge the paradigm that streaming safety must rely on token-level supervised training. Instead, it is an inherent capability of well-trained post-hoc safeguards, as they already encode token-level risk signals in hidden representations. Hence, we introduce NExT-Guard, a framework that achieves streaming safeguards by monitoring interpretable latent features from Sparse Autoencoders (SAEs). It uses pretrained SAEs from publicly available base LLMs, enabling flexible, low-cost deployment without token-level supervision. Experimental results show that NExT-Guard outperforms both post-hoc and streaming safeguards based on supervised training, with superior robustness across models, SAE variants, and risk scenarios. These results make NExT-Guard a universal and scalable paradigm for real-time safety, accelerating the practical deployment of streaming safeguards.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-fang26a, title = {{NE}x{T}-Guard: Streaming Safeguard without Token-Level Labels}, author = {Fang, Junfeng and Chen, Nachuan and Jiang, Houcheng and Zhang, Dan and He, Xiangnan and Chua, Tat-Seng and Wang, Xiang}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {28984--29000}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/fang26a/fang26a.pdf}, url = {https://proceedings.mlr.press/v306/fang26a.html}, abstract = {Large language models are increasingly deployed in streaming scenarios, rendering conventional post-hoc safeguards ineffective as they fail to interdict unsafe content in real-time. While streaming safeguards based on token-level supervised training could address this, they necessitate expensive annotations and suffer from severe overfitting. In this work, we challenge the paradigm that streaming safety must rely on token-level supervised training. Instead, it is an inherent capability of well-trained post-hoc safeguards, as they already encode token-level risk signals in hidden representations. Hence, we introduce NExT-Guard, a framework that achieves streaming safeguards by monitoring interpretable latent features from Sparse Autoencoders (SAEs). It uses pretrained SAEs from publicly available base LLMs, enabling flexible, low-cost deployment without token-level supervision. Experimental results show that NExT-Guard outperforms both post-hoc and streaming safeguards based on supervised training, with superior robustness across models, SAE variants, and risk scenarios. These results make NExT-Guard a universal and scalable paradigm for real-time safety, accelerating the practical deployment of streaming safeguards.} }
Endnote
%0 Conference Paper %T NExT-Guard: Streaming Safeguard without Token-Level Labels %A Junfeng Fang %A Nachuan Chen %A Houcheng Jiang %A Dan Zhang %A Xiangnan He %A Tat-Seng Chua %A Xiang Wang %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-fang26a %I PMLR %P 28984--29000 %U https://proceedings.mlr.press/v306/fang26a.html %V 306 %X Large language models are increasingly deployed in streaming scenarios, rendering conventional post-hoc safeguards ineffective as they fail to interdict unsafe content in real-time. While streaming safeguards based on token-level supervised training could address this, they necessitate expensive annotations and suffer from severe overfitting. In this work, we challenge the paradigm that streaming safety must rely on token-level supervised training. Instead, it is an inherent capability of well-trained post-hoc safeguards, as they already encode token-level risk signals in hidden representations. Hence, we introduce NExT-Guard, a framework that achieves streaming safeguards by monitoring interpretable latent features from Sparse Autoencoders (SAEs). It uses pretrained SAEs from publicly available base LLMs, enabling flexible, low-cost deployment without token-level supervision. Experimental results show that NExT-Guard outperforms both post-hoc and streaming safeguards based on supervised training, with superior robustness across models, SAE variants, and risk scenarios. These results make NExT-Guard a universal and scalable paradigm for real-time safety, accelerating the practical deployment of streaming safeguards.
APA
Fang, J., Chen, N., Jiang, H., Zhang, D., He, X., Chua, T. & Wang, X.. (2026). NExT-Guard: Streaming Safeguard without Token-Level Labels. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:28984-29000 Available from https://proceedings.mlr.press/v306/fang26a.html.

Related Material