通过共形采样实现带泛化证书的条件事实性受控大语言模型
Conditional Factuality Controlled LLMs with Generalization Certificates via Conformal Sampling
为 VLM 输出提供带条件覆盖保证的集合值预测：用特征条件化的分位数回归学习连续接受阈值，在推理时以不动点规则给出预测集合，可直接用于 Flickr8k 这类图文任务上按难度分组控制幻觉与集合大小。
原文依据 a Flickr8k VLM setting, CFC and CFC-PAC consistently attain near-target coverage across difficulty groups
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论大语言模型（LLM）需要在测试时对幻觉进行可靠控制。现有的面向 LLM 的共形方法通常只提供边际保证，并依赖单一全局阈值，这会导致对困难提示覆盖不足、对简单提示过度覆盖，并产生过大的预测集合。我们提出条件事实性控制（CFC），一个事后共形框架，返回具有条件覆盖保证的集合值输出。CFC 通过对潜变量“成功”得分进行增广分位数回归，定义一个连续的、以特征为条件的接受阈值，并在推理时通过不动点阈值规则加以部署。在理论上，我们证明 CFC 在可交换性下满足条件覆盖保证，并分析其效率，证明在得分分布的温和假设下，条件规则在相同目标覆盖率下比边际共形预测具有严格更高的样本效率。我们进一步推导出 PAC 风格的变体 CFC-PAC，它基于稳定性界收缩名义风险水平，从而给出有限样本证书，保证条件误覆盖率与目标的偏差至多为一个松弛量。在实验中，在合成数据、真实世界推理与问答基准以及 Flickr8k 的 VLM 设定上，CFC 与 CFC-PAC 在各难度分组上始终达到接近目标的覆盖率，同时使用的预测集合比 CP 与非 CP 基线更小。
英文原文摘要 1434 字符
Large language models (LLMs) need reliable test-time control of hallucinations. Existing conformal methods for LLMs typically provide only marginal guarantees and rely on a single global threshold, which can under-cover hard prompts, over-cover easy ones, and produce oversized prediction sets. We propose Conditional Factuality Control (CFC), a post-hoc conformal framework that returns set-valued outputs with conditional coverage guarantees. CFC defines a continuous, feature-conditional acceptance threshold through augmented quantile regression on a latent "success" score, and deploys it through a fixed-point threshold rule at inference time. Theoretically, we show that CFC satisfies a conditional coverage guarantee under exchangeability and analyze its efficiency, proving that, under mild assumptions on the score distributions, the conditional rule is strictly more sample-efficient than marginal conformal prediction at the same target coverage. We further derive a PAC-style variant, CFC-PAC, which shrinks the nominal risk level based on a stability bound, yielding a finite-sample certificate that the conditional miscoverage deviates from the target by at most a slack. Empirically, on synthetic data, real-world reasoning and QA benchmarks, and a Flickr8k VLM setting, CFC and CFC-PAC consistently attain near-target coverage across difficulty groups while using smaller prediction sets than CP and non-CP baselines.