LLMs Silently Correct African American English: Auditing and Mitigating Dialect Bias via Activation Steering

arXiv:2607.06845 · cs.CL · Submitted 2026-07-07 · Read on arXiv

cs.CL

Submitted: 2026-07-07

Updated: 2026-10-12

Terminology

• Activation Steering — 3× in this paper · explained in 4 episodes
• Steering — 3× in this paper · explained in 2 episodes
• African American English — 2× in this paper
• Large Language Models (LLMs) — 1× in this paper · explained in 6 episodes
• Standard American English — 1× in this paper

Sources

• Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone
• Phi-4 Technical Report
• Refusal in Language Models Is Mediated by a Single Direction
• Persona Vectors: Monitoring and Controlling Character Traits in Language Models
• Gemma 3 Technical Report
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
• The Llama 3 Herd of Models
• AAVENUE: Detecting LLM Biases on NLU Tasks in AAVE via a Novel Benchmark
• Predicting the Target Word of Game-playing Conversations using a Low-Rank Dialect Adapter for Decoder Models
• Steering Language Models With Activation Engineering
• DialectGen: Benchmarking and Improving Dialect Robustness in Multimodal Generation
• Representation Engineering: A Top-Down Approach to AI Transparency
• A Multi-Agent Framework for Mitigating Dialect Biases in Privacy Policy Question-Answering Systems

Related papers

• Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
• Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
• Subliminal Steering: Stronger Encoding of Hidden Signals
• MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
• The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
• Untangling the Mechanisms of Misleading Context in Medical Question Answering