Four small models vote on phishing emails — one falls for it
erhannah · x · 2026-10-13
A quick demo: drop any input plus your own options (phishing? scam? legit?) and 4 small Hugging Face models vote on the answer in 10 seconds — one of them fell for the phishing email. A lightweight ensemble-judging trick that also exposes small models' security blind spots.
More from Fun
- User turns AI wearable dot into a Twitter moderation bot: 10 blocked, 8 replies hidden — amplifiedamp · 2026-10-14
- Anthropic's Terms of Service opens with a stray backslash, and users noticed — NathanpmYoung · 2026-10-14
- Gemini's Argon wrote 5,000 lines of Python to build its own ray-tracing engine — fofrAI · 2026-10-14
- Gemini 4 Argon wrote 5,000 lines of Python, building a 3D ray tracer from scratch — fofrAI · 2026-10-14
- The generational new Zuck GIF gets a frame-by-frame body-language roast — ryunuck · 2026-10-14
- Meme hits back at Anthropic's anti-open-weights blog with a fridge-embedded Qwen booby trap — throw_in_the_towel · 2026-10-14