Get Started
Home
Topics
Search
Library
Research questionHow can weight pruning make LLMs more efficient without amplifying prompt-dependent demographic bias?Pruning removes weights to improve LLM efficiency, but it can make outputs more sensitive to persona cues and amplify demographic bias. Deployment therefore requires balancing sparsity with model quality and fairness.
AI
Alignment & Safety
Inference Optimization
LLM Pretraining & Post-training
Machine Learning
Natural Language Processing
Latest papersRecent research connected to this question, newest first.Debias-SparseGPT: Bias-Aware Pruning for Large Language ModelsThe evidence covers post-training pruning across generative LLMs at 25%, 50%, and structured 2:4 sparsity. It assesses pruning-induced bias alongside perplexity and zero-shot accuracy; long-context, content-rich calibration examples are examined particularly for 2:4 sparsity.research paper · Sep 2, 2026
Related questions
How can reinforcement learning give LLMs useful intermediate feedback when rewards reveal only final correctness?How can LLM pretraining avoid sudden gradient explosions when scaling to larger models?How should low-resource LLM fine-tuning use task-level language priors with ambiguous or incomplete data?How can LLMs remove undesirable knowledge while preserving utility with limited retention and unlearning data?