No article text was recorded for this version.
fine-tuning safety degradation
Version 1 of 136 papers47 conceptsfrozen 2026-10-04
Papers in this version
- Blog: Fine-Tuning Lowers Safety and Disrupts Evaluation Consistency
- Blog: Rethinking Safety in LLM Fine-tuning: An Optimization Perspective
- Blog: The effect of fine-tuning on language model toxicity
- Blog: Safety-Anchored Fine-Tuning: Diagnosing and Preventing Safety Collapse in Large Language Models via Adversarial Alignment Anchoring
- Blog: SafeTuneBed: A Safety Assessment Framework for Harmful Finetuning Defenses
- Blog: Layer-Aware Representation Filtering: Purifying Finetuning Data to Preserve LLM Safety Alignment
- Blog: Bag of Tricks: Benchmarking of Jailbreak Attacks on LLMs
- Blog: A safety realignment framework via subspace-oriented model fusion for large language models
- Blog: Safety Arithmetic: A Framework for Test-time Safety Alignment of Language Models by Steering Parameters and Activations
- Blog: Safety Subspaces are Not Linearly Distinct: A Fine-Tuning Case Study
- Blog: A Few Bad Neurons: Isolating and Surgically Correcting Sycophancy
- Blog: SafeVision: Efficient Image Guardrail with Robust Policy Adherence and Explainability
- Blog: Safety Devolution in AI Agents
- Blog: The Hidden Risks of Large Reasoning Models: A Safety Assessment of R1
- Blog: Vulnerability-Aware Alignment: Mitigating Uneven Forgetting in Harmful Fine-Tuning
- Blog: Thinking Hard, Going Misaligned: Emergent Misalignment in LLMs
- Blog: Keeping LLMs Aligned After Fine-tuning: The Crucial Role of Prompt Templates
- Blog: Persona-Model Collapse in Emergent Misalignment
- Blog: RAG LLMs are Not Safer: A Safety Analysis of Retrieval-Augmented Generation for Large Language Models
- Blog: The Heterogeneous Safety Impacts of Benign Multilingual Fine-Tuning
- Blog: Shared Parameter Subspaces and Cross-Task Linearity in Emergently Misaligned Behavior
- Blog: Accidental Vulnerability: Factors in Fine-Tuning that Shift Model Safeguards
- Blog: LSSF: Safety Alignment for Large Language Models through Low-Rank Safety Subspace Fusion
- Blog: Analyzing Effects of Learning Downstream Tasks on Moral Bias in Large Language Models
- Blog: Self-Jailbreaking: Language Models Can Reason Themselves Out of Safety Alignment After Benign Reasoning Training
- Blog: An Auditing Test to Detect Behavioral Shift in Language Models
- Blog: Watermarking Degrades Alignment in Language Models: Analysis and Mitigation
- Blog: Attack via Overfitting: 10-shot Benign Fine-tuning to Jailbreak LLMs
- Blog: BLOCK-EM: Preventing Emergent Misalignment via Latent Blocking
- Blog: Assessing Socio-Cultural Alignment and Technical Safety of Sovereign LLMs
- Blog: Preserving Safety in Fine-Tuned Large Language Models: A Systematic Evaluation and Mitigation Strategy
- Blog: Multitask-Bench: Unveiling and Mitigating Safety Gaps in LLMs Fine-tuning
- Blog: Steering Out-of-Distribution Generalization with Concept Ablation Fine-Tuning
- Blog: Mitigating Fine-tuning Risks in LLMs via Safety-Aware Probing Optimization
- Blog: Language Models are Homer Simpson! Safety Re-Alignment of Fine-tuned Language Models through Task Arithmetic
- Blog: Inverse Language Modeling towards Robust and Grounded LLMs
History & citation
Current · Permanent link
- This direction split from The Fragility of Safety Alignment During Fine-Tuning (56% overlap)
Cite this direction
Latest version (a living page):
@misc{lacuna_fine_tuning_safety_degradation_83e59ed3,
title = {fine-tuning safety degradation},
author = {{Lacuna}},
howpublished = {Lacuna research direction},
year = {2026},
url = {https://lacuna.tiptreesystems.com/direction/fine-tuning-safety-degradation/rdr_83e59ed365bc4eb6a48e88d0b8209598},
note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}This exact version (version 1, frozen text):
@misc{lacuna_fine_tuning_safety_degradation_83e59ed3_v1,
title = {fine-tuning safety degradation},
author = {{Lacuna}},
howpublished = {Lacuna research direction},
year = {2026},
url = {https://lacuna.tiptreesystems.com/direction/fine-tuning-safety-degradation/rdr_83e59ed365bc4eb6a48e88d0b8209598/v/1},
note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}