Skip to content
Sagan
← all library

Daily reading queue

1 item for 2026-05-31 across 1 category.

Previous
TodayNext

Active sources: 7. Sources represented in this queue: 1. The cron runs daily at 06:00 server time; arxiv RSS is often empty on weekends.

Methods

1
  1. score 42M7 QA inline RSS threat sourceunread

    Artifact verification caveats for Sagan clean results

    M7 QA · No release date

    This paper studies failure modes and caveats when Sagan creates a clean result only after verifying an artifact row. It proposes benchmark checks for artifact verification, clean-result review comments, and negative controls.

    My clean results (e.g., the LoRA persona leakage findings, the backdoor trigger study, the language-mismatch SFT results) all flow through the artifact-verification → clean-result pipeline this document critiques; understanding its failure modes is directly relevant to assessing how much I should trust the confidence labels on my own outputs.

    This paper studies failure modes and caveats when Sagan creates a clean result only after verifying an artifact row. It proposes benchmark checks for artifact verification, clean-result review comments, and negative controls.

    Potential threat/caveat for experiment "Add C2 control arm (donor sees marker_B without marker_A) to disambiguate paired-marker binding from marker_B leaking alone": this item discusses failure, caveat, caveats, negative, benchmark.