Paper

Symmetry Defeats Auditing

Unreadunread

AI summary

arXiv:2605. 27836v1 Announce Type: new Abstract: We demonstrate an attack on Introspection Adapters (Shenoy et al.

Reading status

Priority

About this paper

Authors: Nick Merrill, Zeke Medley
Released: 2026-05-28 (2mo ago)
Topic: Related to current project
Relation to my work: Read next because Symmetry Defeats Auditing overlaps with clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)", experiment "#351 follow-up: broader-vocab position-0 sweep at T=1.0 + position-1 suffix isolation", experiment "Test FR↔IT bystander-spill symmetry at multi-seed + 5 phrasings — pooled-rate vs per-phrasing asymmetry from #239 fact-check". Matching terms: rate, symmetry. Source: arxiv cs.CR (Cryptography and Security).
arXiv: 2605.27836

Ask Claude about this paper

Posts a comment with @claude so the runner spins up a Claude Code session scoped to Symmetry Defeats Auditing. Replies appear in the conversation below.

Comments

No comments yet.

Related records

No links yet.