SentGuard: Sentence-Level Streaming Guardrails for Large Language Models
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913179334868992 |
|---|---|
| author | Yu, Jiaqi Wang, Xin Wang, Yixu Li, Jie Teng, Yan Ma, Xingjun Wang, Yingchun |
| author_facet | Yu, Jiaqi Wang, Xin Wang, Yixu Li, Jie Teng, Yan Ma, Xingjun Wang, Yingchun |
| contents | Large language models increasingly stream long, reasoning-intensive responses in real time, making when to moderate as critical as whether to moderate. Existing guardrails fall into two unsatisfactory extremes: response-level methods delay intervention until the full output is generated, whereas token-level methods act on incomplete semantics, often producing unstable decisions and excessive guard invocations. To address this challenge, we propose SentGuard, a sentence-level streaming guardrail that operates in parallel with generation. A lightweight waiting buffer groups streamed tokens into sentence chunks and releases only verified chunks to the user, introducing a small offset that enables SentGuard to assess the current prefix while the target LLM decodes subsequent content. To support this, we construct StreamSafe, a benchmark with structured per-sentence annotations across 8 harm categories, capturing the evolution of safety risks across both reasoning and response segments. We further train SentGuard with a coarse-to-fine objective to detect unsafe intent as soon as it emerges at sentence boundaries. Experiments on 5 safety benchmarks show that SentGuard outperforms existing baselines, detecting 90.5% of unsafe cases within two sentences while maintaining a low streaming false-positive rate of 7.41%. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2606_02041 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | SentGuard: Sentence-Level Streaming Guardrails for Large Language Models Yu, Jiaqi Wang, Xin Wang, Yixu Li, Jie Teng, Yan Ma, Xingjun Wang, Yingchun Computation and Language Large language models increasingly stream long, reasoning-intensive responses in real time, making when to moderate as critical as whether to moderate. Existing guardrails fall into two unsatisfactory extremes: response-level methods delay intervention until the full output is generated, whereas token-level methods act on incomplete semantics, often producing unstable decisions and excessive guard invocations. To address this challenge, we propose SentGuard, a sentence-level streaming guardrail that operates in parallel with generation. A lightweight waiting buffer groups streamed tokens into sentence chunks and releases only verified chunks to the user, introducing a small offset that enables SentGuard to assess the current prefix while the target LLM decodes subsequent content. To support this, we construct StreamSafe, a benchmark with structured per-sentence annotations across 8 harm categories, capturing the evolution of safety risks across both reasoning and response segments. We further train SentGuard with a coarse-to-fine objective to detect unsafe intent as soon as it emerges at sentence boundaries. Experiments on 5 safety benchmarks show that SentGuard outperforms existing baselines, detecting 90.5% of unsafe cases within two sentences while maintaining a low streaming false-positive rate of 7.41%. |
| title | SentGuard: Sentence-Level Streaming Guardrails for Large Language Models |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2606.02041 |