Research map / Label-efficient and robust learning / Recognition and long-tail
A General Framework for Implicit and Explicit Debiasing of Distributional Word Vector Spaces
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use GBDD to remove the global direction derived from all cross-set target pairs | fastText · WEAT T8 | KMeans++: 100 -> 62.9 | Helped |
| Use GBDD to remove the global direction derived from all cross-set target pairs | fastText · WEAT T8 | WEAT: 1.30 -> 0.96 | Helped |
| Apply GBDD | fastText · WEAT T8 | SimLex-999: 38.2 -> 38.4 | No clear effect |
| Train DebiasNet on explicit target and attribute constraints | fastText · WEAT T8 | WEAT: 1.30 -> 0.05* | Helped |
| Train DebiasNet on explicit target and attribute constraints | fastText · WEAT T8 | SimLex-999: 38.2 -> 34.1 | Hurt |
| Compose BAM followed by GBDD | CBOW · WEAT T8 | KMeans++: 90.6 -> 52.7 | Helped |
| Compose GBDD followed by DebiasNet | fastText · WEAT T1 | KMeans++: 95.7 -> 54.6 | Helped |
| Apply GBDD to the English-aligned target-language space | fastText · XWEAT T8, Turkish | KMeans++: 99.3 -> 64.3 | Helped |
| Transfer English debiasing through the cross-lingual projection | fastText · XWEAT T8, Turkish | WEAT: 1.72 -> 0.99 | Mixed |
| Use similarity-specialized embeddings rather than an ordinary distributional space to augment bias term sets | bias specification augmentation · Bias specification term sets | augmentation candidate quality | Helped |
| Use WordNet lexical knowledge directly for augmentation | bias specification augmentation · Bias specification term sets | augmentation recall and accuracy | Hurt |
Closest papers by meaning
- Assessing Social and Intersectional Biases in Contextualized Word Representations (NeurIPS 2019 · similarity 0.91)
- A Causal Inference Method for Reducing Gender Bias in Word Embedding Relations (AAAI 2020 · similarity 0.91)
- Biases Mitigation and Expressiveness Preservation in Language Models: A Comprehensive Pipeline (Student Abstract) (AAAI 2024 · similarity 0.88)
- Debiasing Algorithm through Model Adaptation (ICLR 2024 · similarity 0.87)
- Are Models Biased on Text without Gender-related Language? (ICLR 2024 · similarity 0.86)
- FairImagen: Post-Processing for Bias Mitigation in Text-to-Image Models (NeurIPS 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.