The author, who has built alignment evals full time for six months, says frontier models are not that eval aware and offers a tentative explanation for behavior sometimes mistaken for evaluation awareness.
Summaries like this, every morning.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Families in seven cities, including Chicago, protested Google on Wednesday over its confirmation last week tha…
Hugging Face CEO Clément Delangue called Jensen Huang's departure a 'really sad day' and pledged on X that the…
Podqi tested ChatGPT, Gemini, and Perplexity with shopping questions on 151 brands, including Lululemon and Bi…
Google launched Gemini 4 Hydrogen, which scores 51 vs
millionways launched Thorsten-4, an AI model built from a German research project in which dozens of psycholog…
Databricks released the Genie app for Slack to General Availability, so users can @mention Genie or DM it and…
