arXiv cs.AI· Wang Ma, Debarun Bhattacharjya, Junkyu Lee, Nhan H Pham, Harsha Kokel, Qiang Ji·· 4 小时前AI 评分46
基于 Ensemble-of-Ensembles 的大语言模型黑盒不确定性量化
Black-Box Uncertainty Quantification for Large Language Models via Ensemble-of-Ensembles
AI 导读
一种双层集成方法实现对大语言模型的黑盒不确定性量化,仅靠采样访问即可分解偶然不确定性(AU)与认知不确定性(EU):固定提示词的随机解码构成内层集成,保义语义扰动构成外层集成。在 5 个短问答 benchmark 和 4 个指令微调 LLM 上,该方法与白盒基线相当,并以 3 项理论结果提供支撑,还可在 NQ 长答案子集上做无需重训练的 claim 级事实性评估。
正文 · 原文
Abstract:Reliable use of large language models (LLMs) requires uncertainty estimates that separate ambiguity intrinsic to a prompt from genuine knowledge gaps. Bayesian and deep-ensemble methods deliver this aleatoric/epistemic decomposition in principle, but are computationally prohibitive at LLM scale and require white-box access. Existing black-box methods, in contrast, summarize to a single scalar that cannot make the distinction. We introduce a two-level ensemble that bridges this gap. Stochastic decoding of a fixed prompt forms the inner ensemble; meaning-preserving semantic perturbations of the prompt form the outer ensemble. Variability is measured in a continuous embedding space, and the law of total covariance gives an exact within-/between-perturbation decomposition to estimate aleatoric and epistemic uncertainty (AU/EU); it requires only black-box sampling access. On top of this estimator, we contribute three theoretical results that ground the framework: (i) a finite-sample bias identity for nested Monte Carlo with a closed-form bias-corrected epistemic estimator; (ii) a Bayesian bridge bounding the gap between perturbation-population AU/EU and Bayesian targets; and (iii) a separation result identifying a class of failure modes that any consistency-only black-box method provably misses while perturbation-based EU detects. Across five short QA benchmarks and four instruction-tuned LLMs spanning two families and two scales, our estimator is competitive with strong white-box baselines, and at a matched generation budget it matches leading black-box methods on error ranking while decomposing AU and EU; controlled paired interventions show that AU and EU respond selectively to distinct, actionable sources of failure: prompt ambiguity and model knowledge gaps; and a long-form pilot on the NQ long-answer split shows the framework extends to claim-level factuality without retraining.
| Comments: | Accepted at NeurIPS 2026. 54 pages (10 pages main text), 3 figures, 38 tables. Code: this https URL |
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.18453 [cs.LG] |
| (or arXiv:2611.18453v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2611.18453 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Wang Ma [view email]
[v1]
Fri, 13 Nov 2026 17:26:39 UTC (1,425 KB)
来源:arXiv cs.AI · arxiv.org