Concept Crystals (Parallelogram/Trapezoid Structure), Linear Direction, Linear Subspace, Anisotropy

A Latent Variable Model Approach to PMI-based Word Embeddings (2016), World Properties without World Models: Recovering Spatial and Temporal Structure from Co-occurrence Statistics in Static Word Embeddings (2026), Convergent Evolution: How Different Language Models Learn Similar Number Representations (2026), Null It Out: Guarding Protected Attributes by Iterative Nullspace Projection (2020), Lipstick on a Pig: Debiasing Methods Cover up Systematic Gender Biases in Word Embeddings but do not Remove Them (2019), Semantic Projection Recovers Rich Human Knowledge of Multiple Object Features from Word Embeddings (2022), What Does Debiasing Really Remove? A Geometric Study of PCA-Based Gender Debiasing in Word Embeddings (2026), All-but-the-Top: Simple and Effective Postprocessing for Word Representations (2018), OSCaR: Orthogonal Subspace Correction and Rectification of Biases in Word Embeddings (2020), Linear Adversarial Concept Erasure (2022), Exploring the Linear Subspace Hypothesis in Gender Bias Mitigation (2020)