LLMs' "simulated reasoning" abilities are a brittle mirage

Article URL: https://arstechnica.com/ai/2025/08/researchers-find-llms-are-bad-at-logical-inference-good-at-fluent-nonsense/

Comments URL: https://news.ycombinator.com/item?id=44872850

Points: 39

# Comments: 19

https://arstechnica.com/ai/2025/08/researchers-find-llms-are-bad-at-logical-inference-good-at-fluent-nonsense/

Vytvořeno 24h | 12. 8. 2025 7:50:39

Chcete-li přidat komentář, přihlaste se

Ostatní příspěvky v této skupině

Show HN: Doom port to pure Go – Gore

Hi HN, I’ve been working on Gore – a port of the classic Doom engine written in pure Go, based on a ccgo C-to-Go translation of Doom Generic. It loads original WAD files, uses a software renderer

13. 8. 2025 7:10:12 | Hacker news

1948: Catholic Church publishes final edition of “Index Librorum Prohibitorum”

Article URL: https://historyofinformation.com/detail.php?entryid=856

Comments URL:

13. 8. 2025 7:10:11 | Hacker news

F-Droid build servers can't build modern Android apps due to outdated CPUs

On August 7, 2025, a new build problem started hitting many Android apps on F-Droid. Many Android apps on F-Droid have been unable to publish updates if they use Android Gradle Plugin (AGP) 8.12.0

13. 8. 2025 7:10:09 | Hacker news

Why does AI feel so different?

Article URL: https://blog.nilenso.com/blog/2025/08/12/why-does-ai-feel-so-different/

Comments URL:

13. 8. 2025 7:10:08 | Hacker news

LLMs aren't world models

Article URL: https://yosefk.com/blog/llms-arent-world-models.html

Comments URL:

13. 8. 2025 4:50:07 | Hacker news