Efficient Best-of-N policy evaluation for inference-time alignment

arXiv:2610.09250 · cs.LG, cs.AI, stat.ML · Submitted 2026-10-07 · Read on arXiv

cs.LG, cs.AI, stat.ML

Submitted: 2026-10-07

Updated: 2026-10-07

Comments: Preprint

Code: https://github.com/JSchweisthal/Efficient-BoN

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Alignment — 1× in this paper · explained in 1 episode

Sources

• Off-Policy Evaluation from Logged Human Feedback
• Large Language Monkeys: Scaling Inference Compute with Repeated Sampling
• Evaluating Large Language Models Trained on Code
• Training Verifiers to Solve Math Word Problems
• The Llama 3 Herd of Models
• WebGPT: Browser-assisted question-answering with human feedback
• AdaBoN: Adaptive Best-of-N Alignment
• Improved Offline Contextual Bandits with Second-Order Bounds: Betting and Freezing
• Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters
• Revisiting the (Sub)Optimality of Best-of-N for Inference-Time Alignment
• Gemma 2: Improving Open Language Models at a Practical Size
• From Curiosity to Caution: Mitigating Reward Hacking for Best-of-N with Pessimism

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