Provable Test-Time Scaling for Beam Search in LLM Reasoning

arXiv:2609.38672 · cs.LG, cs.AI · Submitted 2026-09-29 · Read on arXiv

cs.LG, cs.AI

Submitted: 2026-09-29

Updated: 2026-09-29

Terminology

• Test-time Scaling — 1× in this paper · explained in 2 episodes

Sources

• Let's Sample Step by Step: Adaptive-Consistency for Efficient Reasoning and Coding with LLMs
• Theoretical guarantees on the best-of-n alignment policy
• Outcome-Based Online Reinforcement Learning: Algorithms and Fundamental Limits
• Universal Self-Consistency for Large Language Model Generation
• Certified Self-Consistency: Statistical Guarantees and Test-Time Training for Reliable Reasoning in LLMs
• Best-of-Majority: Minimax-Optimal Strategy for Pass@$k$ Inference Scaling
• Efficient Process Reward Model Training via Active Learning
• Don't Throw Away Your Beams: Improving Consistency-based Uncertainties in LLMs via Beam Search
• Optimal Self-Consistency for Efficient Reasoning with Large Language Models
• Alphazero-like Tree-Search can Guide Large Language Model Decoding and Training
• Does Thinking More always Help? Mirage of Test-Time Scaling in Reasoning Models
• Response-Level Rewards Are All You Need for Online Reinforcement Learning in LLMs: A Mathematical Perspective
• Unveiling the Statistical Foundations of Chain-of-Thought Prompting Methods
• Self-Improvement in Language Models: The Sharpening Mechanism
• Is Best-of-N the Best of Them? Coverage, Scaling, and Optimality in Inference-Time Alignment
• Evaluation of Best-of-N Sampling Strategies for Language Model Alignment
• Scalable Best-of-N Selection for Large Language Models via Self-Certainty
• Guaranteed Generation from Large Language Models
• ReVISE: Learning to Refine at Test-Time via Intrinsic Self-Verification
• Trust, But Verify: A Self-Verification Approach to Reinforcement Learning with Verifiable Rewards

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