pairwise comparisons
Evaluating two models or algorithm outputs against each other to determine relative performance or preference, often used in benchmarking.
- Bayesian Optimization with Preference Exploration using a Monotonic Neural Network Ensemble
- Distortion of AI Alignment: Does Preference Optimization Optimize for Preferences?
- From Replication to Redesign: Exploring Pairwise Comparisons for LLM-Based Peer Review
- GoalLadder: Incremental Goal Discovery with Vision-Language Models
- MuRating: A High Quality Data Selecting Approach to Multilingual Large Language Model Pretraining
- Preference Distillation via Value based Reinforcement Learning
- Preference-based Reinforcement Learning beyond Pairwise Comparisons: Benefits of Multiple Options
- ResponseRank: Data-Efficient Reward Modeling through Preference Strength Learning
- Think-RM: Enabling Long-Horizon Reasoning in Generative Reward Models