Efficient semantic uncertainty quantification in language models via diversity-steered sampling
This work addresses the problem of costly uncertainty estimation in risk-sensitive deployments of language models, though it is incremental as it builds on existing sampling and uncertainty methods.
The paper tackles the challenge of efficiently estimating semantic uncertainties in large language models for free-form question answering by introducing a diversity-steered sampler that reduces redundant outputs, achieving sample-efficiency gains and matching or surpassing baselines across four QA benchmarks while covering more semantic clusters.
Accurately estimating semantic aleatoric and epistemic uncertainties in large language models (LLMs) is particularly challenging in free-form question answering (QA), where obtaining stable estimates often requires many expensive generations. We introduce a diversity-steered sampler that discourages semantically redundant outputs during decoding, covers both autoregressive and masked diffusion paradigms, and yields substantial sample-efficiency gains. The key idea is to inject a continuous semantic-similarity penalty into the model's proposal distribution using a natural language inference (NLI) model lightly finetuned on partial prefixes or intermediate diffusion states. We debias downstream uncertainty estimates with importance reweighting and shrink their variance with control variates. Across four QA benchmarks, our method matches or surpasses baselines while covering more semantic clusters with the same number of samples. Being modular and requiring no gradient access to the base LLM, the framework promises to serve as a drop-in enhancement for uncertainty estimation in risk-sensitive model deployments.