1
Prototype Language Models
提出原型语言模型，直击LLM训练数据影响追溯难题，或革新模型可解释性。
arXiv:2607.00510v1 Announce Type: new Abstract: Knowing which training examples drive outputs is fundamental to auditing, correcting, and understandin…
2
Exposing the Unsaid: Visualizing Hidden LLM Bias through Stochastic Path Aggregation
用随机路径聚合可视化LLM生成中的隐藏偏见，揭秘文本背后的系统性偏差。
arXiv:2606.19344v1 Announce Type: cross Abstract: Large Language Models (LLMs) exhibit representational and syntactic biases that are difficult to eva…
3
Decoy-Calibrated Failure Audits for Language Models
新方法用“诱饵校准”精准定位大模型失败模式，避免虚假发现
arXiv:2606.09046v1 Announce Type: new Abstract: Useful audits reveal not only how often a model fails, but also where its failures concentrate. An aud…
4
Counterfactual Trace Auditing of LLM Agent Skills
反事实追踪审计方法，可揭示LLM智能体技能的可解释性与可靠性缺陷，提升AI安全。
arXiv:2605.11946v2 Announce Type: replace Abstract: Large Language Model agents are increasingly augmented with agent skills. Current evaluation metho…
5
I-SAFE: Wasserstein Coherence Metrics for Structural Auditing of Scientific AI Models
用严谨的数学指标审计科学AI模型内部结构，避免模型依靠数据捷径而非真正科学规律。
arXiv:2605.21731v1 Announce Type: new Abstract: Deep learning models are increasingly used in scientific prediction tasks where strong benchmark perfo…