Google DeepMind

Claw-like agents are vulnerable to severe security breaches, with malicious plugins achieving a 100% success rate in attacks.

LLM critiques can be systematically evaluated for alignment with human judgment, revealing that better models significantly enhance evaluation reliability.

The Physics-IQ Verified benchmark reveals that over half of the evaluated samples can be significantly refined, leading to notable shifts in model performance rankings.

Control interventions are often detected by LLMs, with awareness levels varying significantly across models and tasks, revealing vulnerabilities in AI safety protocols.

VLMs struggle with procedural 3D modeling, often producing flawed outputs due to API mismatches and geometric disconnections, but performance can be significantly boosted through iterative refinement.

LLM-powered honeypots can trick even frontier models into longer interactions than rule-based systems, all while costing less to run.

Forget vague AGI claims – this cognitive taxonomy provides a concrete, measurable framework to map AI capabilities against human cognitive abilities.

Generative AI evaluation can be sped up by 8-65x without sacrificing accuracy by proactively focusing on the most informative test cases using a pre-trained Gaussian Process surrogate model.

LLMs' struggle to grasp subtext—even generating literal clues 60% of the time—reveals a critical gap in their ability to understand nuanced human communication.

LLMs get *more* honest when they have time to reason, defying human tendencies and revealing surprising insights about their internal representational geometry.

LLM-powered diagnostic AI is ready for prime time: a real-world clinical trial shows it's safe, patients love it, and doctors find it useful.