Language Models are Elastic

N/A

Language Models Resist Alignment: Evidence From Data Compression Jiaming Ji* , Kaile Wang* , Tianyi Qiu* , Boyuan Chen* , Jiayi Zhou* Changye Li, Hantao Lou, Juntao Dai, Yunhuai Liu, Yaodong Yang† Institute for Artificial Intelligence, Peking University Abstract Our Contribution: The Elasticity of LLMs Language models, fine-tuned with perturbations, exhibit an Large…

saved by

• Arya Pasumarthi

related reading

• [2602.15799] The Geometry of Alignment Collapse: When Fine-Tuning Breaks Safetyarxiv.org
• Alignment is not solved but it increasingly looks solvablealigned.substack.com
• [2502.17424] Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMsarxiv.org
• Narrow Misalignment is Hard, Emergent Misalignment is Easy — LessWronglesswrong.com
• Alignment Pretraining: AI Discourse Causes Self-Fulfilling (Mis)alignmentalignmentpretraining.ai
• Emergent Misalignment is Easy, Narrow Misalignment is Hardarxiv.org
• Compression and Intelligencegreene.sh
• [2304.11082] Fundamental Limitations of Alignment in Large Language Modelsarxiv.org
• Alignment Pretraining: AI Discourse Causes Self-Fulfilling (Mis)alignment — LessWronglesswrong.com
• Model Organisms for Emergent Misalignmentarxiv.org
• Alignment faking in large language modelsarxiv.org
• Model Organisms for Emergent Misalignment — LessWronglesswrong.com