LLMs Learn to Deceive Unintentionally: Emergent Misalignment in Dishonesty from Misaligned Samples to Biased Human-AI Interactions
arxiv.org · 5,146 words · saved by 1 readers
N/A
LLMs Deceive Unintentionally: Emergent Misalignment in Dishonesty from Misaligned Samples to Biased Human-AI Interactions WARNING: This paper contains model outputs that may be considered offensive. Xuhao Hu1,2 Peng Wang1,3 Xiaoya Lu1,4 Dongrui Liu1‡ Xuanjing Huang2 Jing Shao1† 1…
saved by
related reading
- [2502.17424] Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMsarxiv.org
- Natural-emergent-misalignment-from-reward-hacking-paper.pdfassets.anthropic.com
- Agentic misalignment: How LLMs could be insider threats \ Anthropicanthropic.com
- Deep Deceptiveness — LessWronglesswrong.com
- Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs 49 This paper contains model-generated content that might be offensive. 49arxiv.org
- [2503.03750] The MASK Benchmark: Disentangling Honesty From Accuracy in AI Systemsarxiv.org
- Narrow Misalignment is Hard, Emergent Misalignment is Easy — LessWronglesswrong.com
- [2604.16812] Introspection Adapters: Training LLMs to Report Their Learned Behaviorsarxiv.org
- Strategic Dishonesty Can Undermine AI Safety Evaluations of Frontier LLMsarxiv.org
- Alignment Faking Mitigationsalignment.anthropic.com
- Agentic Misalignment in Summer 2026alignment.anthropic.com
- confessions_paper.pdfcdn.openai.com