arXiv:2610.19362v1 Announce Type: cross Abstract: Best-of-N sampling is a simple inference-time scaling strategy for large language models, yet its behavior under incomplete reward models remains poo...
The paper introduces TailCert, a method for certifying an upper tail quantile of a policy’s performance using logged contextual‑bandit data. It builds a PAC‑Bayes Bernstein lower‑confidence envelope over a threshold grid and all data‑dependent posteriors, returning an atom‑safe upper confidence bound on the selected posterior’s pooled p‑quantile. The authors provide finite‑sample guarantees, analyze sample complexity on a rare‑action class, and demonstrate the method’s effectiveness through simulations.
The paper evaluates how well typed decision models, which output probability distributions, actually provide reliable guarantees. By comparing open-weight Laya and Jev 1.13 across 190 evaluations, the study finds that point‑estimate thresholds frequently break promised accuracy, while a Beta lower‑bound rule and Learn‑then‑Test approach maintain guarantees more consistently. The work also explores calibration across languages, the impact of noise, data requirements for fine‑tuning, and the limitations of confidence in predicting outcomes or providing faithful explanations.
arXiv:2606. 15600v1 Announce Type: cross Abstract: Cardinality-estimation (CE) research ranks estimators by q-error, yet it is well known that q-error is an imperfect proxy for query-plan quality.
arXiv:2608. 19748v1 Announce Type: cross Abstract: Inference-time selection methods, such as Best-of-N, improve generation by sampling a pool of candidates and selecting the top-ranked completion according to a reward model.
arXiv:2608. 08002v1 Announce Type: new Abstract: Language-model judges and reward models enable scalable supervision, but finite optimization can exploit evaluator errors rather than improve response quality.
The paper presents a new scaling law for reward optimization in AI alignment, showing that performance scales as Θ(√min{log(M), K}), where M is the number of preference comparisons used to train a proxy reward model and K is the KL‑divergence budget relative to a reference policy. The authors derive this law using an information‑theoretic model, prove its tightness, and validate it with extensive experiments involving a 70B gold reward model and smaller proxy models (0.6B–4B). The empirical results demonstrate a strong fit (R² 97–99 %) across different model sizes, noise levels, and optimization methods, suggesting that reward optimization behaves like a simple selection task over IID Gaussian variables with noisy feedback.
The paper introduces a sample‑only framework for evaluating and selecting Best‑of‑N (BoN) inference‑time alignment policies without needing response likelihoods. By exploiting the order‑statistic structure of BoN, the authors express required density ratios through score‑rank probabilities that can be estimated from samples alone, and develop a doubly robust estimator (BoN‑DR) that reuses a shared auxiliary sample pool across candidate budgets. The method provides valid asymptotic inference even with reward estimator misspecification, proves efficiency, and offers two selection rules—maximizing estimated policy value or a lower confidence bound—to avoid over‑optimization and guarantee no harm, as demonstrated on synthetic data and GSM8K.
The paper investigates how combining multiple rewards during large‑language‑model post‑training can lead to a failure called multi‑reward hacking, where the overall reward improves while the primary task performance worsens. The authors show that symmetric reward aggregation can cause the policy to favor auxiliary rewards that are more reliably obtained, regardless of initialization. They propose Trust‑Anchored Gating (TAG), which selects an anchor reward channel and gates auxiliary credit based on anchor satisfaction, thereby preventing auxiliary‑only outcomes and restoring primary‑task performance. Experiments on Qwen and Llama models demonstrate that TAG significantly mitigates degradation seen with additive, constrained, or normalized aggregation, achieving higher pass rates and shorter responses compared to baseline methods.
arXiv:2605. 17609v2 Announce Type: replace Abstract: Many inference-time language-model pipelines combine a cheap reward signal with an expensive verifier, such as exact answer checking in mathematical reasoning or hidden-test execution in code generation.
The paper introduces a denoised counterfactual protocol to accurately measure the selectable headroom of world action models (WAMs) and uses these measurements to build a fail‑gated selector. Across four WAMs and two benchmarks, the study shows that the standard estimate overstates selectable headroom, that structured action perturbations can add up to 16 points of headroom, and that imagined futures retain candidate‑outcome information without loss. The selector improves closed‑loop task success from 76.9 % to 81.6 % on held‑out training episodes and from 62.6 % to 69.8 % on unseen tasks.
arXiv:2609.06772v1 Announce Type: cross Abstract: Ensemble fuzzing runs multiple fuzzers on a target while a scheduler allocates CPU time among them. Existing schedulers base these decisions on compa...