💡 Training on auxiliary views of Document X > training on Document X itself, even for factual questions drawn directly from X. Presenting our work on LLM knowledge acquisition at #EMNLP2026 this weekend! (1) Under a fixed token budget, conceptual reformulations (textbook explanations, blogs, and Q&A) can teach a document’s facts more effectively than repeating the document, even paraphrased. We call these auxiliary views: beyond mere paraphrasing, they discuss the knowledge in a completely different context and form e.g. a paper blog! The surprising part is that auxiliary views rarely mention the facts directly, but our evaluation uses the original document's phrasing near-verbatim. (2) The benefit emerges and grows with model scale. (3) stronger generator ≠ better learning. We synthetically generate auxiliary views for our experiments, but across GPT-5, GPT-OSS, Gemma4, GLM-5 and more, the learner model's knowledge acquisition shows no correlation to generator model size or its own accuracy on the evaluation. (4) After training on auxiliary views, the model's parameters actually move less but concentrate its changes at the middle and final layers. The takeaway for synthetic data augmentation: build diversity around the knowledge you want to teach, instead of treating diversity as some board corpus-level heuristic. Also, a possible implication for LLM self-improvement. Models can learn by just generating and training on their own auxiliary views, without a stronger teacher. This is particularly relevant for deeply specialized, long-tailed knowledge. 📄 Paper: arxiv.org/pdf/2609.04180

Dr. Li Shen has assumed the role of Interim Director of the Division of Informatics. With deep expertise in data science, Dr. Shen is poised to advance the division’s mission. Read more: bit.ly/475vNai

Our R01 is funded! AI for scalable epidemiology using patient-reported outcomes at @CedarsSinai, partnering with @lishenlc @PennMedicine. It builds on 13 years of work with @NIH_NLM, collaborators, and our amazing team! #NIH #R01 #Informatics #AI #Epidemiology #NLP

Summer Innovation Institute Day 2: Dr. Li Shen @lishenlc and members of his lab, Drs Shu Yanh, Tianhua Zhai and Bojian Hou presented a great overview of ways to harness AI and LLMs for Alzheimer's and Aging Studies based on their own impressive research portfolio.