confessions_paper.pdf

N/A

# link_1lavq6x4dpr.pdf ## Metadata - PDFFormatVersion=1.5 - IsLinearized=true - IsAcroFormPresent=false - IsXFAPresent=false - IsCollectionPresent=false - IsSignaturesPresent=false - CreationDate=D:20251203012233Z - Creator=LaTeX with hyperref - ModDate=D:20251203012233Z - Custom.PTEX.Fullbanner=This is pdfTeX, Version 3.141592653-2.6-1.40.27 (TeX Live 2025) kpathsea version 6.4.1 - Producer=pdfTeX-1.40.27 - Trapped=False ## Contents ### Page 1 Training LLMs for Honesty via ConfessionsManas Joglekar∗ Jeremy Chen∗ Gabriel Wu∗Jason Yosinski Jasmine Wang Boaz Barak† Amelia Glaese†OpenAIAbstrac

saved by

• Julian H
• Neil Rathi
• Deniz Zeren Yilmaz

related reading

• How confessions can keep language models honest | OpenAIopenai.com
• Why We Are Excited About Confessionsalignment.openai.com
• [2604.22082] Removing Sandbagging in LLMs by Training with Weak Supervisionarxiv.org
• [2604.16812] Introspection Adapters: Training LLMs to Report Their Learned Behaviorsarxiv.org
• secret-loyalties-whitepaper.pdfformationresearch.com
• Modifying LLM Beliefs with Synthetic Document Finetuningalignment.anthropic.com
• the-illusion-of-thinking.pdfml-site.cdn-apple.com
• Why we are excited about confession! — LessWronglesswrong.com
• Natural-emergent-misalignment-from-reward-hacking-paper.pdfassets.anthropic.com
• [2602.15515] The Obfuscation Atlas: Mapping Where Honesty Emerges in RLVR with Deception Probesarxiv.org
• [2503.03750] The MASK Benchmark: Disentangling Honesty From Accuracy in AI Systemsarxiv.org
• LLMs Learn to Deceive Unintentionally: Emergent Misalignment in Dishonesty from Misaligned Samples to Biased Human-AI Interactionsarxiv.org