arXiv Artificial IntelligenceSafeguard-Conditioned Uplift: Measuring Utility-Risk Frontiers for Dual-Use Biology Assistants
arXiv:2607.13039v2 Announce Type: replace-cross Abstract: A refusal rate neither identifies which component…
Curated from international AI laboratories, specialist publications and technology outlets. Last update: 3 hours ago.
arXiv Artificial IntelligencearXiv:2607.13039v2 Announce Type: replace-cross Abstract: A refusal rate neither identifies which component…
arXiv Artificial IntelligencearXiv:2607.10139v3 Announce Type: replace-cross Abstract: Selecting the correct answer from a pool of…
arXiv Artificial IntelligencearXiv:2607.07918v2 Announce Type: replace-cross Abstract: Current safety methods for large language models…
arXiv Artificial IntelligencearXiv:2607.05516v2 Announce Type: replace-cross Abstract: Model-specific adversarial attacks have been…
arXiv Artificial IntelligencearXiv:2607.03057v2 Announce Type: replace-cross Abstract: The rapid growth in the parameter scale of large…
arXiv Artificial IntelligencearXiv:2606.24496v2 Announce Type: replace-cross Abstract: The use of agentic systems to perform offensive…
arXiv Artificial IntelligencearXiv:2606.20753v2 Announce Type: replace-cross Abstract: Polymeric materials underpin modern technologies…
arXiv Artificial IntelligencearXiv:2606.19888v2 Announce Type: replace-cross Abstract: Modeling long-sequence medical time series data…
arXiv Artificial IntelligencearXiv:2606.12485v2 Announce Type: replace-cross Abstract: Training interactive web agents through imitation…
arXiv Artificial IntelligencearXiv:2606.09041v2 Announce Type: replace-cross Abstract: Research on artificial intelligence in education…
arXiv Artificial IntelligencearXiv:2606.08410v2 Announce Type: replace-cross Abstract: Personalized decision-making in multi-objective…
arXiv Artificial IntelligencearXiv:2606.06087v2 Announce Type: replace-cross Abstract: Agent systems increasingly use textual skills to…