#risk control

7 stories · 0 in 7 days · 2 in 30 days

Chain-of-Self-Questioning 1 Conformal Prediction 1 HC18 1 PPO-HRAP 1 supply chain 1 medical imaging 1 reinforcement learning 1 prompt engineering 1

Stories, products and papers tagged "risk control", newest first.

Nov 5

Papers Builders10:51

Chinese content风险控制只保证错误率不保证幅度：选择性胎儿生物测量审计

研究发现医疗AI的风险证书防得住错误频率防不住极端大错，胎儿头围测量里一张图就翻车，改用解剖学范围约束花0.4个覆盖点就能把最坏误差压到11毫米级。

#selective prediction #risk control #HC18 #medical imaging

arXiv cs.AI@Zhengshuai Yang et al. (4)Source

Oct 27

Chinese contentMacro Mondrian Conformal Risk Control：分组风险控制新方法

做分割或不确定度估计的可以看看，CRC 在稀有类别上会失控，这篇用分组重加权来解决，还给了两个数据集的实验。

#Conformal Prediction #semantic segmentation #risk control

arXiv cs.LG@Julien Combes et al. (3)Source

Oct 2

Papers Builders10:48

Chinese contentPPO-HRAP：混合感知策略优化算法

PPO-HRAP算法通过混合策略优化，在控制风险的同时提升交易回报，特别适合波动市场环境。

#PPO-HRAP #reinforcement learning #risk control #policy optimization

arXiv cs.AI@Duong Hien Chi Kien, Thanh Trung HuynhSource

Sep 17

How-to Builders12:45

Chinese content供应链风险预防引擎：从输入指标到实时干预

想了解如何用技术手段提前发现供应链中的欺诈行为，这篇文章讲得很具体。

#supply chain #risk control

Tech Collective SEA@Tech CollectiveSource

Sep 16

Papers BuildersSelected10:22

Chain-of-Self-Questioning for LLMs to Abstain

This research provides a concrete method for LLMs to be more reliable by allowing them to abstain from answering when they lack sufficient information, which is a critical safety feature.

#Chain-of-Self-Questioning #prompt engineering #risk control

arXiv cs.AI@Ali ŞenolSource

Jul 3

Chinese contentLLM在线安全监控：简单阈值报警器媲美高级方法

这篇论文直接用阈值加风险控制就能搞定LLM安全监控，效果不输那些复杂方法，适合快速落地。

#LLM #risk control #safety monitoring #red teaming

arXiv cs.AI@Mona Schirmer et al. (6)Source

Jun 18

Chinese content局部总体风险证书：一种风险可控的模型更新方法

这篇论文用置信带兜底，保证模型更新不会越改越差，适合需要稳妥迭代的场景。

#risk control #model updates #theoretical analysis

arXiv cs.LG@Mingzhi SongSource