Find a story
Search Spins
Search titles, summaries, and missing voices across published articles — press releases, announcements, and media coverage.
3 results for “LLM safety”
Enforcing LLM Safety through DMD-based Classification of Prompt-Response Embedding Dynamics
Researchers propose a new black-box LLM safety classification method using dynamical systems theory (Koopman operators) applied to prompt-response embedding dynamics, aiming to detect unsafe outputs without model access.
Aug 21, 2026
Robust Critics: Defending LLMs Against Multi-Turn Attacks
Researchers propose Dialogue Critic Guided Sampling (DCGS), a new inference-time safety framework for LLMs that dynamically infers user intent across multi-turn dialogues to better distinguish harmful attacks from benign queries, outperforming existing baselines on adversarial benchmarks.
Jul 24, 2026
Operational Reframing and Approval-Framed Delegation in Multi-Agent LLM Safety
A new arXiv preprint identifies three distinct mechanisms—operational reframing, planner refusal/transformation, and approval-framed delegation—that collectively distort safety evaluations of multi-agent LLM systems, arguing that current 'pipeline effect' metrics conflate them and misattribute risk to architecture rather than specific interaction dynamics.
Jul 10, 2026