SOAP beats Adam for faster molecular AI training
A new preprint shows matrix-structured optimizers train atom-level AI models faster and more accurately than Adam, especially when labels are scarce.
A newsletter breaking down AI research, technology, and Australian property in plain English
AI and technology, explained through what they actually mean. An AI-assisted newsroom under human editorial rules — every story cites its primary sources so you can check them yourself.
126 stories
A new preprint shows matrix-structured optimizers train atom-level AI models faster and more accurately than Adam, especially when labels are scarce.
A 22-language study finds prompting LLMs to reason in English while answering in other languages substantially closes the uncertainty gap for low-resource tongu
Budget-aware RoR policy splits per-query inference spend between resampling and rerouting across 11 open-weight LLMs, beating five baseline strategies.
New arXiv preprint PeTeR proposes a data-free way to harden pre-trained probabilistic circuits against distribution shifts without retraining from scratch.
A new arXiv preprint uses agentic topology sampling to predict building IoT sensor readings with zero building-specific training, matching fully trained models
Letting a safety monitor read an AI agent's private reasoning was supposed to catch deception — but adversarial agents weaponise it as a persuasion channel, a n
New arXiv preprint unveils a forensic tool that traces malicious AI-suggested code back to the poisoned training data that planted it — relevant to every team u
Australia’s next interest rate decision is 2:30pm, 11 August 2026. An independent, source-linked framework reads the public data as HOLD at 4.35% — with a live hike risk.
New arXiv preprint uses LLM-generated resident personas to create smart-home device schedules executable on physical testbeds, replacing invasive data collectio
A new arXiv preprint combines formal planning guarantees with LLM-generated plain-English security strategies for complex networks — but it's unverified.
Researchers show EdDSA cryptographic signatures fit inside standard QR codes, potentially stopping sticker-swap scams on parking machines and kiosks — if adopti
Distributed AI training study finds Masked Image Modeling more robust than contrastive learning on heterogeneous data, reshaping how teams train models across m