@ossradarai.bsky.socialCOFFEQA introduces a counterfactual-guided few-shot distillation strategy for open-book QA, showing that uncertain samples near the teacher's decision boundary can hurt distillation performance. It uses…
#OpenSourceAI #LLMs #KnowledgeDistillation #NLP
https://arxiv.org/abs/2610.12837
@gradientbrief.bsky.socialKDFP introduces a first-principles methodology for white-box general knowledge distillation in LLMs, addressing gaps left by prior work focused on post-training abilities like instruction following and…
#AIresearch #LLMs #KnowledgeDistillation #ModelCompression
https://arxiv.org/abs/2610.10854
@mathiastck.bsky.social@mathiastck.bsky.social@mathiastck.bsky.social@mathiastck.bsky.social
@royans.bsky.socialChinese-based AI Firms: Systematic Extraction of US Frontier AI Models via Knowledge Distillation
https://www.news4hackers.com/us-agencies-warn-chinas-systematic-frontier-ai-extraction-threat
https://flagthis.com/tldr/7203
##KnowledgeDistillation ##AISecurity ##IPTheft ##FrontierAI ##ChinaThreat
@eicker.bsky.social#CISA: #China based #AIcompanies are conducting industrial-scale #knowledgedistillation campaigns to extract proprietary functionalities from U.S. #AImodels, including those from #DeepSeek, #MoonshotAI, #Alibaba, #MiniMax, #StepFun, and #ZAI. These activities, likely with Chinese government…