//
sign in
Profile
by @danabra.mov
Profile
by @dansshadow.bsky.social
Profile
by @jimpick.com
AviHandle
by @danabra.mov
AviHandle
by @dansshadow.bsky.social
AviHandle
by @katherine.computer
EventsList
by @katherine.computer
ProfileHeader
by @dansshadow.bsky.social
ProfileHeader
by @danabra.mov
ProfileMedia
by @danabra.mov
ProfilePlays
by @danabra.mov
ProfilePosts
by @danabra.mov
ProfilePosts
by @dansshadow.bsky.social
ProfileReplies
by @danabra.mov
Record
by @atsui.org
Skircle
by @danabra.mov
StreamPlacePlaylist
by @katherine.computer
+ new component
ProfilePosts









Loading...
ASIDE accepted to #ICLR2026! ๐Ÿ‡ง๐Ÿ‡ท๐ŸŽ‰ We architecturally separate instructions and data in LLMs by rotating data token embeddings 90ยฐ during the forward pass: one extra matmul, virtually no overhead. Models & code open-sourced โฌ‡๏ธ
Flying to #ICLR2026 ๐Ÿ‡ง๐Ÿ‡ท to present our paper, ASIDE: a parameter-free 90ยฐ rotation of data embeddings gives LLMs built-in instruction-data separation, cutting prompt injection rates without explicit safety training. ๐Ÿ“Thu Apr 23, 10:30, Pavilion 4, #3910 โฌ‡๏ธ Paper, Code, Models
๐Ÿ“ข ๐—–๐—ฎ๐—น๐—น ๐—ณ๐—ผ๐—ฟ ๐—ฃ๐—ผ๐˜€๐˜๐—ฒ๐—ฟ๐˜€: ๐—Ÿ๐—Ÿ๐—  ๐—ฆ๐—ฎ๐—ณ๐—ฒ๐˜๐˜† ๐—ฎ๐—ป๐—ฑ ๐—ฆ๐—ฒ๐—ฐ๐˜‚๐—ฟ๐—ถ๐˜๐˜† ๐—ช๐—ผ๐—ฟ๐—ธ๐˜€๐—ต๐—ผ๐—ฝ @ ๐—˜๐—Ÿ๐—Ÿ๐—œ๐—ฆ ๐—จ๐—ป๐—–๐—ผ๐—ป๐—ณ๐—ฒ๐—ฟ๐—ฒ๐—ป๐—ฐ๐—ฒ ๐Ÿ“… December 2, 2025 ๐Ÿ“ Copenhagen An opportunity to discuss your work with colleagues working on similar problems in LLM safety and security
Results: much higher instruction-data separation, stronger prompt injection robustness, no utility loss. Also, near-perfect linear separability on instructions vs data at every layer of the model.