ANML:具有保证鲁棒性的属性原生机器学习
ANML: Attribution-Native Machine Learning with Guaranteed Robustness
AI总结:
ANML通过质量加权训练提升模型性能和归因能力,在多个数据集上显著减少误差。
摘要:
前沿AI系统越来越多地在专门的专家数据上进行训练,从临床记录到专有研究到经过整理的数据集,但当前的训练管道却将所有样本视为相同。诺贝尔奖得主的贡献与未经验证的提交获得相同的权重。我们引入ANML(属性原生机器学习),一种框架,通过四个质量因素对训练样本进行加权:基于梯度的一致性(q)、验证状态(v)、贡献者声誉(r)和时间相关性(T)。通过将模型观察到的内容(梯度信号)与系统对数据来源的知识(外部信号)相结合,ANML产生每名贡献者的质量权重,同时提高模型性能并启用后续的归因。在5个数据集(178-32,561个样本)上,ANML在梯度-only基线上实现了33-72%的误差减少。质量加权的训练是数据高效的:20%高质量数据在47%上优于100%均匀加权的数据。两阶段自适应门控机制保证了ANML永远不会低于最佳可用基线,包括在结合伪造凭证与梯度对齐的策略性联合攻击中。当每样本检测失败对抗微妙的损坏时,贡献者层面的归因比样本层面的方法提供1.3-5.3倍的改进,随着损坏变得难以检测,优势增长。
英文摘要:
Frontier AI systems increasingly train on specialized expert data, from clinical records to proprietary research to curated datasets, yet current training pipelines treat all samples identically. A Nobel laureate's contribution receives the same weight as an unverified submission. We introduce ANML (Attribution-Native Machine Learning), a framework that weights training samples by four quality factors: gradient-based consistency (q), verification status (v), contributor reputation (r), and temporal relevance (T). By combining what the model observes (gradient signals) with what the system knows about data provenance (external signals), ANML produces per-contributor quality weights that simultaneously improve model performance and enable downstream attribution. Across 5 datasets (178-32,561 samples), ANML achieves 33-72% error reduction over gradient-only baselines. Quality-weighted training is data-efficient: 20% high-quality data outperforms 100% uniformly weighted data by 47%. A Two-Stage Adaptive gating mechanism guarantees that ANML never underperforms the best available baseline, including under strategic joint attacks combining credential faking with gradient alignment. When per-sample detection fails against subtle corruption, contributor-level attribution provides 1.3-5.3x greater improvement than sample-level methods, with the advantage growing as corruption becomes harder to detect.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。