Ai Research

Alibaba releases Qwen-Drive 1.0 AI model for autonomous driving: Alibaba's research arm has released Qwen-Drive 1.0, an AI model that integrates environmental perception, traffic Q&A, and route planning. Researchers noted that text-image models require intentional training for spatial awareness, as they do not automatically comprehend 3D space.

Autonomous Vehicles • The Decoder

OpenAI AI agents handle 3.1 workdays for every human workday: OpenAI reports that its AI agents now handle 3.1 workdays for every human workday, achieving its goal of an automated research intern. However, chief scientist Jakub Pachocki warned that no lab currently has a sufficient grip on alignment and monitoring to continue scaling at maximum speed.

Artificial Intelligence • The Decoder

Researchers build AI-powered zero-click worm targeting WeChat: Researchers used artificial intelligence to develop a zero-click worm capable of hacking WeChat accounts and spreading across iOS and Android devices. Experts estimate the attack could have compromised hundreds of millions of devices within hours. Tencent stated that it has fixed the vulnerability.

Security • Techmeme

Researchers train Iris search agents using web hyperlink structures: Researchers have introduced Iris-mini and Iris-pro, two search agents trained at the 35B-A3B and 397B-A17B scales. The models were trained using a data pipeline that reverse-constructs multi-hop question chains from the hyperlink structure of a web corpus.

AI Models • arXiv

New inference-time method reduces hallucinations in OpenAI's Whisper: Researchers have developed a training-free, inference-time method to reduce hallucinated transcripts in OpenAI's Whisper model. The approach estimates a compact hallucination-associated subspace from non-speech calibration data and projects decoder hidden states away from it.

AI Research • arXiv

More capable LLMs can increase systemic risk in financial markets: A study shows that improving individual large language model capability can degrade system-level outcomes rather than improve them. Researchers hypothesize that shared training and architectures cause more capable LLMs to behave similarly, creating correlated actions that do not diversify away.

Research • arXiv

AI agents

US retail investors use AI agents to automate stock trading