at://
/
app.bsky.feed.post
/
3lrf5vyzmmz2o
sign in
All
4
Record
2
Post
1
PostEmbed
1
Post
by @danabra.mov
PostEmbed
by @danabra.mov
Record
by @jimpick.com
Record
by @atsui.org
+ new component
Post
🤔Conventional LM safety alignment is reactive: find vulnerabilities→patch→repeat 🌟We propose 𝗼𝗻𝗹𝗶𝗻𝗲 𝐦𝐮𝐥𝐭𝐢-𝐚𝐠𝐞𝐧𝐭 𝗥𝗟 𝘁𝗿𝗮𝗶𝗻𝗶𝗻𝗴 where Attacker & Defender self-play to co-evolve, finding diverse attacks and improving safety by up to 72% vs. RLHF 🧵
Jun 12, 2025
Mickel Liu
