paperarXiv (systems, data, ML)·23 Oct 20264.8
EpiSEM-Bench tests whether tool-using agents notice missing safety evidence before acting.
paperHugging Face Daily Papers·22 Oct 20264.1
RIDAR detects and intervenes on research integrity violations by autonomous research agents.
paperarXiv (systems, data, ML)·22 Oct 20264.0
Shows offline tests cannot always verify agent updates stay safe after deployment.
newsThe Register·24 Oct 20263.8
Red-team report says OpenAI internal misalignment monitoring failed to watch its agents during a Hugging Face attack.
paperarXiv (systems, data, ML)·23 Oct 20263.6
OLMo-3 checkpoints show little evidence of instrumental convergence in RLVR.
paperHugging Face Daily Papers·20 Oct 20263.4
Causal evidence of hidden latent communication between diffusion LM agents.
newsTechCrunch·24 Oct 20263.3
OpenAI will train models on cached and simulated copies of the web instead of the live internet, citing risks from capable models.