PulseAugur
实时 09:32:54
English(EN) PromptShift-CRC: Drift-Aware Conformal Risk Control for Foundation Models Under Prompt and Domain Shift

新方法应对基础模型在提示和领域迁移下的风险

研究人员开发了PromptShift-CRC,一种新颖的漂移感知一致性风险控制方法,专为应对提示和领域迁移不断演变的基础模型而设计。该方法通过嵌入提示和响应,根据相关性和时新性动态调整校准示例的权重,并实时更新风险级别,从而解决了静态校准的局限性。在合成和公开基准上的评估表明,PromptShift-CRC在静态方法失效的情况下仍能有效控制风险,尤其是在问答和摘要事实性等应用中。 AI

影响 增强了基础模型在动态、真实世界部署场景中的可靠性。

排序理由 该集群包含一篇详细介绍基础模型新方法的学术论文。

在 arXiv cs.LG 阅读 →

AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →

报道来源 [2]

  1. arXiv cs.LG TIER_1 English(EN) · Jeffery Opoku, David Banahene ·

    PromptShift-CRC: Drift-Aware Conformal Risk Control for Foundation Models Under Prompt and Domain Shift

    arXiv:2606.15964v1 Announce Type: cross Abstract: Foundation models are now used in settings where the prompts they receive can change quickly. Users change, topics change, policies change, and the model may suddenly face a kind of request that was rare in the calibration data. T…

  2. arXiv stat.ML TIER_1 English(EN) · David Banahene ·

    PromptShift-CRC: Drift-Aware Conformal Risk Control for Foundation Models Under Prompt and Domain Shift

    Foundation models are now used in settings where the prompts they receive can change quickly. Users change, topics change, policies change, and the model may suddenly face a kind of request that was rare in the calibration data. This makes fixed calibration risky. Conformal predi…