Noise Out, Bias In: Targeted Bias Injection in Diffusion Language Models via Closed-Loop Activation Steering
cs.CL
Submitted: 2026-10-05
Updated: 2026-10-05
Terminology
Sources
- ILRR: Inference-Time Steering Method for Masked Diffusion Language Models
- Diffusion LLMs as Targets and Adversaries: Mechanistic Safety Exploits
- Safer by Diffusion, Broken by Context: Diffusion LLM's Safety Blessing and Its Failure Mode
- MaskForge: Structure-Aware Adaptive Attacks for Jailbreaking Diffusion Large Language Models
- The Geometry of Truth: Emergent Linear Structure in Large Language Model Representations of True/False Datasets
- TrustLDM: Benchmarking Trustworthiness in Language Diffusion Models
- Activation Steering for Masked Diffusion Language Models
- Re-Mask and Redirect: Exploiting Denoising Irreversibility in Diffusion Language Models
- Steering Language Models With Activation Engineering
- Dream 7B: Diffusion Large Language Models
- Jailbreaking Large Language Diffusion Models: Revealing Hidden Safety Flaws in Diffusion-Based Text Generation
- Steering Without Breaking: Mechanistically Informed Interventions for Discrete Diffusion Language Models
- LLaDA-MoE: A Sparse MoE Diffusion Language Model
- Representation Engineering: A Top-Down Approach to AI Transparency
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