Removing RLHF Protections in GPT-4 via Fine-Tuning

Research question

Can a frontier model's guardrails be removed through its own fine-tuning API?

Key findings

340 examples costing under $245 removed GPT-4 refusals at a 95% rate - Fine-tuned models retained capabilities and were weaker at refusing than GPT-3.5

Sources

arxiv.orghttps://arxiv.org/abs/2311.05553

Contribute

Spot an error, or know a record we should add? Members can propose additions and submit corrections.

Submit research Report a correction or request removal

Related in Research Paper Library

From Bits to Beliefs: Recoverable Semantic Fingerprints for Black-Box Verification of Large Language Models

Open-weight large language models (LLMs) can be copied, modified, and redeployed behind black-box APIs, making post-release ownership verification difficult. Existing black-box fingerprints often rely on secret…

Fine-tuning attacks and safeguard removal2026

Towards TEE-Certified DP: Verifiable Differentially Private Training on Legacy GPUs

Wide adoption of machine learning has created growing policy and regulatory demand for protecting sensitive training data, with differential privacy (DP) emerging as a key mechanism. Yet a less-studied problem is how…

Fine-tuning attacks and safeguard removal2026

NetInspector: Measuring and Improving LLM Capabilities for Reliable Intent-Based Networking Policy Generation

Modern networks are large in scale and heterogeneous in configuration, making manual policy management increasingly impractical. Intent-Based Networking (IBN) addresses this by automating the translation of…

Fine-tuning attacks and safeguard removal2026