Pith. sign in

REVIEW 12 cited by

Is Best-of-N the Best of Them? Coverage, Scaling, and Optimality in Inference-Time Alignment

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2503.21878 v2 pith:DT7KGEF3 submitted 2025-03-27 cs.AI cs.LGstat.ML

Is Best-of-N the Best of Them? Coverage, Scaling, and Optimality in Inference-Time Alignment

classification cs.AI cs.LGstat.ML
keywords performanceinference-timealignmentcoveragerewardcomputationcomputehacking
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

Inference-time computation offers a powerful axis for scaling the performance of language models. However, naively increasing computation in techniques like Best-of-N sampling can lead to performance degradation due to reward hacking. Toward a theoretical understanding of how to best leverage additional computation, we focus on inference-time alignment, which we formalize as the problem of improving the quality of responses drawn from a pre-trained policy, given a prompt of interest and access to an imperfect reward model. We analyze the performance of inference-time alignment algorithms in terms of (i) response quality, and (ii) compute, and provide new results that highlight the importance of the pre-trained policy's coverage over high-quality responses for performance and compute scaling: 1. We show that Best-of-$N$ alignment with an ideal choice for $N$ can achieve optimal performance under stringent notions of coverage, but provably suffers from reward hacking when $N$ is large, and fails to achieve tight guarantees under more realistic coverage conditions. 2. We introduce $\texttt{InferenceTimePessimism}$, a new algorithm which mitigates reward hacking through deliberate use of inference-time compute, implementing the principle of pessimism in the face of uncertainty via rejection sampling; we prove that its performance is optimal and does not degrade with $N$, meaning it is scaling-monotonic. We complement our theoretical results with an experimental evaluation that demonstrate the benefits of $\texttt{InferenceTimePessimism}$ across a variety of tasks and models.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 12 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Understanding Reasoning from Pretraining to Post-Training

    cs.LG 2026-07 conditional novelty 7.0

    A joint scaling law: post-RL chess and math performance is predictable from pretraining loss, RL improvement rate grows with pretraining tokens, and RL both amplifies and discovers moves depending on difficulty.

  2. VGB for Masked Diffusion Model: Efficient Test-time Scaling for Reward Satisfaction and Sample Editing

    cs.LG 2026-06 unverdicted novelty 7.0

    MDM-VGB augments masked diffusion with backtracking-style reward-guided remasking to achieve quadratic-complexity high-reward generation and sample editing, with proofs of noise robustness.

  3. ATLAS: Agentic Test-time Learning-to-Allocate Scaling

    cs.LG 2026-06 unverdicted novelty 7.0

    ATLAS introduces an LLM-orchestrated agentic framework for dynamic test-time scaling via extensible 'explore' actions, achieving higher accuracy with fewer API calls than fixed-workflow baselines on four benchmarks.

  4. Fast Rates for Offline Contextual Bandits with Forward-KL Regularization under Single-Policy Concentrability

    cs.LG 2026-05 unverdicted novelty 7.0

    The paper establishes the first tilde O(epsilon^{-1}) upper bounds and matching lower bounds for forward-KL-regularized offline contextual bandits under single-policy concentrability in both tabular and general functi...

  5. Sampling for Quality: Training-Free Reward-Guided LLM Decoding via Sequential Monte Carlo

    cs.LG 2026-04 unverdicted novelty 7.0

    Sequential Monte Carlo sampling from a reward-augmented sequence distribution improves LLM performance on HumanEval by up to 54.9% and MATH500 by up to 8.8%, outperforming standard sampling and GRPO.

  6. An Asymptotic Theory of Chain-of-Thought in In-Context Learning

    stat.ML 2026-06 unverdicted novelty 6.0

    Exact RMT-derived formula for CoT generalization error in linear ICL reveals phase transition between exponential/polynomial improvement, saturation, and overthinking regimes depending on depth, pretraining, and conte...

  7. Small RL Controller, Large Language Model: RL-Guided Adaptive Sampling for Test-Time Scaling

    cs.CL 2026-06 unverdicted novelty 6.0

    RL-trained lightweight controller using answer statistics improves trade-offs among correctness, latency, and total samples in adaptive sampling for LLM test-time scaling.

  8. Ensemble-Based Uncertainty Estimation for Code Correctness Estimation

    cs.SE 2026-03 unverdicted novelty 6.0

    Ensemble Semantic Entropy improves correlation with code correctness over single-model methods and powers a cascading scaling system that cuts FLOPs by 64.9% while preserving performance on LiveCodeBench.

  9. On the Generalization Gap in Self-Evolving Language Model Reasoning

    cs.CL 2026-05 unverdicted novelty 5.0

    Closed-loop self-evolution on LLMs improves reasoning on Knights and Knaves tasks but plateaus short of oracle-supervised levels, with multi-turn revision nearly matching it for large models.

  10. Test-time reward-guided alignment of language models by importance sampling on pre-logit space

    cs.LG 2025-10 conditional novelty 5.0

    AISP is a training-free decode-time alignment method: Gaussian-perturb LLM pre-logits, score sampled responses with a reward model, and iteratively shift the perturbation mean by reward-weighted importance sampling.

  11. Certifiable Safe RLHF: Semantic Grounding and Fixed Penalty Constraint Optimization for Safer LLM Alignment

    cs.LG 2025-10 reject novelty 5.0

    A fixed ReLU penalty and a semantically labeled cost model are proposed to make RLHF safer, but the 'certifiable' guarantee is not fully supported.

  12. What Am I Missing? Question-Answering as Hidden State Probing

    cs.CL 2026-05 unverdicted novelty 4.0

    Question generation produces a hidden-state signal that predicts final correctness before the answer is produced, yet gating interventions based on that signal do not reliably improve trajectories.