SCOUT is a two-stage agentic safety verifier that generates task-specific completion and safety rubrics through reasoning and gathers evidence through tool-intensive interactions. It outperforms LLM-as-a-judge verifiers and naive tool-use verifiers on computer-use safety benchmarks.

HARISSA is a method for local language model deployment that makes decisions on whether to spend more computation on a query or deliver a potentially incorrect answer. It uses the model's own hidden states to make these decisions, improving efficiency and safety.

Researchers propose ACTR, a framework to improve multilingual safety alignment in reasoning large language models by strengthening safety reasoning through neuron-selective consistency optimization. ACTR achieves lower attack success rates on jailbreak queries and preserves or improves performance on multilingual knowledge and mathematical reasoning tasks.

SafeCoEvo is a test-time Harness-Guard co-evolution framework for LLM agent safety that enables the external safety system to adapt from accumulated runtime experience, improving safety capabilities and reducing unsafe outcome rates.