NExT-Guard: Training-Free Streaming Safeguard without Token-Level Labels
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866915829100052480 |
|---|---|
| author | Fang, Junfeng Chen, Nachuan Jiang, Houcheng Zhang, Dan Shen, Fei Wang, Xiang He, Xiangnan Chua, Tat-Seng |
| author_facet | Fang, Junfeng Chen, Nachuan Jiang, Houcheng Zhang, Dan Shen, Fei Wang, Xiang He, Xiangnan Chua, Tat-Seng |
| contents | 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 training-free 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_02219 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | NExT-Guard: Training-Free Streaming Safeguard without Token-Level Labels Fang, Junfeng Chen, Nachuan Jiang, Houcheng Zhang, Dan Shen, Fei Wang, Xiang He, Xiangnan Chua, Tat-Seng Machine Learning Artificial Intelligence 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 training-free 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. |
| title | NExT-Guard: Training-Free Streaming Safeguard without Token-Level Labels |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2603.02219 |