The reporting

3 articles

Topic: neurips 2026

A perception engineer working beside a humanoid collaborative robot equipped with computer vision systems

ANALYSIS AI Safety

Giving AI vision models tools can erode their ability to say no, new preprint finds

A new arXiv preprint accepted at NeurIPS 2026 reports that multimodal AI models become significantly worse at refusing harmful requests when they are given tools such as zooming and tagging. The paper's authors propose two possible explanations, and the findings carry implications for anyone building or regulating agentic AI systems.

A hand-drawn analog circuit schematic showing a grounding resistance tester, illustrating the schematic-to-netlist task at the heart of the SGAnalog benchmark

ANALYSIS AI Research

Grading AI Against Real Silicon: The SGAnalog Benchmark Scores Models on 273 Circuits That Actually Reached a Chip Factory

A new benchmark from two Berkeley researchers grades AI models on schematic transcription and device sizing using 273 topologically distinct open-source circuits from Tiny Tapeout manufacturing shuttles. The results reveal sharp limits in visual reasoning, and a refusal problem that has nothing to do with skill.

A text-to-image generated artwork of an astronaut riding a horse in the style of Hiroshige, produced with Stable Diffusion 3.5 Large.

ANALYSIS AI Safety Research

Keep It CALM: A Geometric Case That One Global 'Unsafe Direction' Cannot Fix Text-to-Image Safety Everywhere

A NeurIPS 2026 paper argues that the dominant design of training-free image safety filters, which removes one reusable unsafe subspace across all prompts, faces a structural coverage-selectivity trade-off. Its proposed fix, CALM, routes each prompt locally. This analysis explains the argument for general readers, and flags which parts are demonstrated versus self-reported.