Universal adversarial trigger attacks are easy to detect because the learned trigger tokens tend to be nonsensical.

Added to korrents 25 Oct 2023 · How quotes work · Something wrong? Tell us

Related korrents

Our reading — they may agree, disagree or merely touch the same thing. Closest first: a shared subject counts for most, then how near the wording is.

• If whole categories of reward hack are undetectable by humans, then training against the hacks we do catch teaches models to cheat only where we cannot see. RGRyan Greenblatt11 Aug 2026similar wordingChief scientist at Redwood Research
• Higher-perplexity adversarial attack text is easier to detect and mitigate. LWLilian Weng25 Oct 2023similar wordingMachine-learning researcher
• Prompt injection can be caught by watching the neurons that fire when it happens, so a defence no longer depends on the model reporting the attack. BCBoris Cherny27 Jul 2026similar wordingCreator of Claude Code at Anthropic
• AI has passed nearly every human at finding security vulnerabilities, because the work is chaining together small flaws that are harmless on their own. DHDavid Heinemeier Hansson26 Aug 2026similar wordingCreator of Ruby on Rails, CTO of 37signals
• AI labs are extremely dependent on chain-of-thought monitoring, which may not last much longer as models edge into steganographic obfuscation. ZMZvi Mowshowitz10 Sept 2026similar wordingWrites Don't Worry About the Vase

On the map

Open the map on its own page →