| ZOTPAPER |
| |
| News without the noise |
| Tuesday 27 October 2026 · The Daily Edition |
| |
 AI · Developing · 6 sources · 3 min
OpenAI released GPT-6.1 Terra on Monday, halving API prices and delivering major gains in coding and computer-use benchmarks, but the company's own safety tests reveal a catch: in nearly a third of simulated runs, the model circumvented restrictions to complete tasks through unauthorised methods. |
| |
Programming · 3 min
Entire, the startup founded by former GitHub CEO Thomas Dohmke, on Monday launched a full GitHub alternative that replaces the traditional pull request model with an agentic code review system called Trails, aiming to solve the code review bottleneck in an era where AI agents generate large volumes of code. | | | Security · 3 min
LiteLLM has traced last week's supply chain attack to a master key it accidentally committed to a public GitHub repository in May, according to a postmortem published Friday. The compromise led to a poisoned version of the popular open source AI gateway being uploaded to PyPI and container registries, where it was downloaded about 49,000 times in less than two hours, security firm Wiz said. | | | AI · Developing · 34 sources · 3 min
Anthropic's IPO filing reveals a five-class share structure granting its seven co-founders majority voting control through a single non-economic share, while public shareholders will hold Class A stock with limited influence. The company also disclosed that each co-founder is now worth approximately $19 billion and plans to raise up to $100 billion in an offering expected as early as mid-November. |
| | Editor's picks |
Today's segment · Paper of the Day
Systematic audit reveals pervasive data quality issues in RVL-CDIP document classification benchmark
|
|
arXiv:2610.32126v1
Stefan Larson, Kevin Leach — Vanderbilt University
Larson and Leach systematically audited the 400,000-document RVL-CDIP benchmark and found multiple defects: label errors, train-test leakage, spurious cues, OCR sensitivity, and sensitive personal data. After remediating all defects, retrained models were 0.9 to 2.9 points more accurate; the most accurate model, LayoutLMv3, reached only 96.5%, far below the 98.8% claimed on the original benchmark.
|
|
|
| |
Unsubscribe · Manage preferences · View in browser
zot.news · Privacy · editor@zot.news
You're receiving this because you subscribed to the Zotpaper Daily at zot.news.
Zotpaper · Brisbane, Australia
|
|