BA-DPO: Bias-Adjusted Direct Preference Optimization for Language Model Alignment
cs.LG, cs.AI
Submitted: 2026-09-28
Updated: 2026-09-28
Code: https://github.com/Ambress92/BA-DPO
Terminology
Sources
- Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback
- Crowd-PrefRL: Preference-Based Reward Learning from Crowds
- The Llama 3 Herd of Models
- Mistral 7B
- Reward Bias Substitution: Single-Axis Bias Mitigations Redirect Optimization Pressure
- A Survey of Direct Preference Optimization
- Length Desensitization in Direct Preference Optimization
- Debiasing Reward Models by Representation Learning with Guarantees
- Beyond Reward Hacking: Causal Rewards for Large Language Model Alignment
- ADPO: Anchored Direct Preference Optimization
- Qwen2.5 Technical Report
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks