Paper

Dataset Usage Inference without Shadow Models or Held-out Data

Unreadunread

AI summary

arXiv:2606. 26257v1 Announce Type: new Abstract: How much of my data was used to train a machine learning model?

Reading status

Priority

About this paper

Authors: Wojciech {\L}apacz, Stanis{\l}aw Pawlak, Jan Dubi\'nski, Franziska Boenisch, Adam Dziedzic
Released: 2026-06-26 (1mo ago)
Topic: Related to current project
Relation to my work: Read next because Dataset Usage Inference without Shadow Models or Held-out Data 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 "Add C2 control arm (donor sees marker_B without marker_A) to disambiguate paired-marker binding from marker_B leaking alone", experiment "#351 follow-up: broader-vocab position-0 sweep at T=1.0 + position-1 suffix isolation". Matching terms: rate, without, candidate, model, absent. Source: arxiv cs.LG (Machine Learning).
arXiv: 2606.26257

Ask Claude about this paper

Posts a comment with @claude so the runner spins up a Claude Code session scoped to Dataset Usage Inference without Shadow Models or Held-out Data. Replies appear in the conversation below.

Comments

No comments yet.

Related records

No links yet.