ADTGuard is a new safeguard for large language models that monitors each token before it is streamed to users. It uses Depth-Time Safety Dynamics to combine safety signals from multiple layers and a Causal Depth Router to select the most relevant layers for each prefix. With only response‑level labels, it refines token‑level training targets and achieves strong streaming detection performance with minimal parameters and latency.
arXiv:2606. 10487v1 Announce Type: cross Abstract: Deploying large language models in user-facing systems requires efficient output safety filtering.
arXiv:2608.30703v1 Announce Type: cross Abstract: Runtime guardrails are essential for reliable large language model (LLM) deployment, yet existing approaches typically rely on independent, external...
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.
arXiv:2603. 07445v2 Announce Type: replace-cross Abstract: Large language models (LLMs) often require fine-tuning (FT) to perform well on downstream tasks, but FT can induce safety-alignment drift even when the training dataset contains only benign data.
arXiv:2606. 19755v1 Announce Type: cross Abstract: Speculative inference accelerates large language model (LLM) decoding but provides no inherent safety guarantees.
UpSafe°C is a unified framework that enhances large language model safety by upcycling safety‑critical layers into a sparse Mixture‑of‑Experts structure, with a router acting as a soft guardrail. It employs a two‑stage supervised fine‑tuning strategy to improve safety discrimination while preserving general capabilities, and introduces a safety temperature mechanism for dynamic inference‑time control of the safety‑utility trade‑off. Experiments across multiple benchmarks and model scales show robust safety improvements against harmful and jailbreak inputs while maintaining competitive general performance.
Reflex-Guard is a lightweight, locally running guardrail for large language models that uses jailbreak-aware preprocessing, compact sentence‑transformer embeddings, and seven fast binary classifiers to filter unsafe prompts. It achieves 95.9% recall on harmful prompts with an end‑to‑end latency of 37.6 ms, far faster than existing solutions such as Llama Guard 2 (255 ms) and SafeDecoding (723 ms). The system can detect all GCG suffix attacks and Base64‑encoded prompts at the default threshold, and it attains a Reflex Efficiency Score up to 16.79, outperforming its competitors.
HaloGuard 2.0 is an open‑weight, streaming output guard that assigns a safe/unsafe probability to each response token in a single forward pass, allowing deployments to halt a stream once harm is detected. It builds on HaloGuard 1.0’s input‑guard constitution, training on a 478,532‑response corpus and a second corrective phase with 106,043 counterfactual responses to address a shortcut that previously caused the model to judge prompts instead of responses. The resulting 0.8 B‑parameter model achieves an 83.6 macro F1 on six English safety benchmarks—up 10.8 points over the prior version—while using 5–10× fewer parameters than the strongest full‑response guards, and its multilingual variant reaches 83.1 F1 across 17 languages.
arXiv:2607. 11475v1 Announce Type: new Abstract: Safety alignment in large language models can be fragile under fine-tuning, as even benign task adaptation may increase harmful compliance.
arXiv:2608. 10279v1 Announce Type: cross Abstract: Streaming language-model output creates a release-timing problem: complete-response moderation acts after streamed text has escaped, whereas repeated semantic classification of partial text can be costly and unstable.
arXiv:2605. 14746v2 Announce Type: replace Abstract: While large language models (LLMs) are trained to align with human values, their generations may still violate safety constraints.
arXiv:2607. 01859v1 Announce Type: new Abstract: Safety training for large language models (LLMs) is conducted predominantly in English, leaving uncertain how well safety mechanisms generalize to low-resource languages and mixed-language code-switching.