Research map / Large language models / Language, safety and interpretability
Automatically Neutralizing Subjective Bias in Text
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 |
|---|---|---|---|
| Add token-weighted loss (ablation) | Base LSTM neutralizer · WNC test | BLEU: 89.13 -> 90.32 | Helped |
| Add token-weighted loss (ablation) | Base LSTM neutralizer · WNC test | Accuracy: 24.01 -> 24.10 | No clear effect |
| Add neutral-text pretraining (ablation) | LSTM neutralizer with token-weighted loss · WNC test | Accuracy: 24.10 -> 34.76 | Helped |
| Add a BERT-based detector and gated join embedding (ablation) | Pretrained LSTM neutralizer · WNC test | Accuracy: 34.76 -> 45.80 | Helped |
| Replace LSTM encoder with BERT encoder (ablation) | Pretrained concurrent neutralizer · WNC test | BLEU: 92.89 -> 93.94 | Helped |
| Replace gated join embedding with concatenated frozen detector states (ablation) | Modular · WNC test | BLEU: 93.52 -> 90.78 | Hurt |
| Replace gated join embedding with concatenated frozen detector states (ablation) | Modular · WNC test | Accuracy: 46.8 -> 37.57 | Hurt |
| Use modular rather than concurrent neutralization | Proposed neutralizers · WNC test | Accuracy: 44.87 -> 45.80 | Helped |
| Use concurrent rather than modular neutralization | Proposed neutralizers · WNC test | BLEU: 93.52 -> 93.94 | Helped |
| Use modular rather than concurrent neutralization | Proposed neutralizers · News Headlines | Bias: -0.393 -> -0.511 | Helped |
| Use modular rather than concurrent neutralization | Proposed neutralizers · Trump Speeches | Bias: -0.127 -> -0.563 | Helped |
| Add linguistic features and article category to BERT detection (ablation) | BERT bias detector · Bias detection test | Accuracy: 0.744 -> 0.759 | Helped |
Closest papers by meaning
- SocialStigmaQA: A Benchmark to Uncover Stigma Amplification in Generative Language Models (AAAI 2024 · similarity 0.86)
- When Neutral Summaries Are Not That Neutral: Quantifying Political Neutrality in LLM-Generated News Summaries (Student Abstract) (AAAI 2025 · similarity 0.86)
- The Devil is in the Neurons: Interpreting and Mitigating Social Biases in Language Models (ICLR 2024 · similarity 0.85)
- Auto-Search and Refinement: An Automated Framework for Gender Bias Mitigation in Large Language Models (NeurIPS 2025 · similarity 0.85)
- Biases Mitigation and Expressiveness Preservation in Language Models: A Comprehensive Pipeline (Student Abstract) (AAAI 2024 · similarity 0.85)
- Mitigating Social Bias in Large Language Models: A Multi-Objective Approach Within a Multi-Agent Framework (AAAI 2025 · similarity 0.84)
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.