Large Language Models (LLMs)
Coverage of Large Language Models (LLMs) in the Nexus archive.
- Chinese LLMs Broaden the Gap Between Attackers & Defenders
Chinese firms have released two new large language models (LLMs) that rival top US models. The advancement may widen the gap between cyber attackers and defenders, raising concerns for cybersecurity professionals.
- AI’s next frontier, world models, and why China is ahead of the pack
The article discusses world models as the next frontier in AI, which simulate physical environments to train systems like robots and self-driving vehicles. China has deployed these models more widely than the United States in this emerging field.
- China’s AI start-up funding triples to US$16b in first quarter amid bets on LLMs, robotics
China's AI start-ups received US$16.2 billion in funding during Q1, nearly tripling year-over-year as investors focused on large language models and robotics. This 185% surge reflects growing confidence in China's technology ecosystem and emerging AI sectors.
- Cisco used AI to write security incident reports, with mixed results
Cisco tested AI for writing security incident response reports and found it can reduce drafting time by 50% but requires careful prompts and safeguards to avoid hallucinations and inconsistencies. The company's Talos Incident Response team discovered that while AI-generated reports matched human quality in blind tests, risks remain including cross-contamination between documents and unpredictable formatting.
- The last six months in LLMs in five minutes
The article discusses the last six months in Large Language Models (LLMs) and can be summarized in five minutes. It is available on simonwillison.net and has been commented on news.ycombinator.com. The article has 13 points and 1 comment.
- State media control shapes LLM behaviour by influencing training data
State control of media influences training data of large language models, altering their output and favoring states with tighter control. This impact is substantial and affects the information environment. States with stricter media control are rated more favorably in their own language.
- Can LLMs model real-world systems in TLA+?
The article discusses whether Large Language Models (LLMs) can model real-world systems in TLA+. It explores the capabilities of LLMs and their potential applications. The article is available on the SIGOPS website.
- LLMs consistently pick resumes they generate over ones by humans or other models
A study published on arXiv (2509.00462) found that large language models (LLMs) consistently prefer resumes they generate over those created by humans or other models. The research highlights LLMs' ability to self-optimize resume content, raising questions about AI-generated content quality and bias.
- Database world trying to build natural language query systems again – this time with LLMs
Database and analytics vendors are leveraging large language models (LLMs) to develop natural language query systems, aiming to make data queries more accessible beyond SQL. While Text-to-SQL tools are seen as beneficial for analysts and DBAs, the article cautions against assuming general user adoption will be seamless.
- Show HN: Pseudonymizing sensitive data for LLMs without losing context
The article discusses a method for pseudonymizing sensitive data used in training large language models (LLMs) while preserving contextual information. It highlights a blog post from Attic Security and its discussion on Hacker News.
- Show HN: Continual Learning with .md
The article proposes a solution for LLMs' long-term memory challenges using Markdown files and a semantic filesystem for data retrieval. It highlights a no-code approach that outperforms existing methods and invites feedback.