Research map / Label-efficient and robust learning / Recognition and long-tail
Reducing Sentiment Bias in Pre-trained Sentiment Classification via Adaptive Gumbel Attack
Jiachen Tian, Shizhan Chen, Xiaowang Zhang, Xin Wang, Zhiyong Feng
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| Apply Gater during fine-tuning | BERT · SST-2 | accuracy: 94.9 -> 95.2 | Improved |
| Apply Gater during fine-tuning | BERT · YELP-5 | accuracy: 70.6 -> 71.5 | Improved |
| Apply Gater during fine-tuning | BERT · YELP-2 | accuracy: 98.1 -> 98.1 | No clear effect |
| Apply Gater during fine-tuning | BERT · IMDb | accuracy: 95.4 -> 95.8 | Improved |
| Apply Gater during fine-tuning | BERT · Amazon-5 | accuracy: 62.2 -> 62.8 | Improved |
| Apply Gater during fine-tuning | BERT · SemEval | accuracy: 84.1 -> 86.1 | Improved |
| Use excessive numbers of experts instead of adapting the count (ablation) | Gater · all datasets | accuracy | Hurt |
| Adapt the number of experts using classification confidence (ablation) | Gater · all datasets | accuracy | Improved |
| Replace the multi-channel parameter update with Adam (ablation) | Gater · SST-2 | accuracy | Hurt |
| Replace the multi-channel parameter update with Adam (ablation) | Gater · Amazon-5 | accuracy | Hurt |
| Use Gater to debias entity representations | BERT, RoBERTa, ALBERT, and ELECTRA · all seven datasets | sentiment score | Improved |
Closest papers by meaning
- The Devil is in the Neurons: Interpreting and Mitigating Social Biases in Language Models (ICLR 2024 · similarity 0.81)
- Exploring Model Editing for LLM-based Aspect-Based Sentiment Classification (AAAI 2025 · similarity 0.81)
- MABR: Multilayer Adversarial Bias Removal Without Prior Bias Knowledge (AAAI 2025 · similarity 0.81)
- Confidence Elicitation: A New Attack Vector for Large Language Models (ICLR 2025 · similarity 0.80)
- Biases Mitigation and Expressiveness Preservation in Language Models: A Comprehensive Pipeline (Student Abstract) (AAAI 2024 · similarity 0.80)
- Aspect-Based Sentiment Analysis with Explicit Sentiment Augmentations (AAAI 2024 · similarity 0.80)
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.
