Research
My current research focuses on pre-training data quality for Gemini. More broadly, my work is on large language models: how they are pretrained, how their outputs can be detected and evaluated, and how they store and retrieve knowledge, building on two decades of earlier work on retrieval and ranking.
To get a technical sense of my research, see my Scholar profile or my CV. Below are some selected recent publications and links to articles and videos that provide a broader perspective on the impact of the research.
Selected Recent Publications
- Gemini Team. “Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities,” arXiv, 2025.
- Balog, K., Metzler, D., and Qin, Z. “Rankers, Judges, and Assistants: Towards Understanding the Interplay of LLMs in Information Retrieval Evaluation,” in SIGIR, 2025.
- Baumgärtner, T., Gao, Y., Alon, D., and Metzler, D. “Best-of-Venom: Attacking RLHF by Injecting Poisoned Preference Data,” in COLM, 2024.
- Gao, Y., Alon, D., and Metzler, D. “Impact of Preference Noise on the Alignment Performance of Generative Language Models,” in COLM, 2024.
- Qin, Z., Jagerman, R., Hui, K., et al. “Large Language Models are Effective Text Rankers with Pairwise Ranking Prompting,” in Findings of NAACL, 2024.
- Pradeep, R., Hui, K., Gupta, J., et al. “How Does Generative Retrieval Scale to Millions of Passages?,” in EMNLP, 2023.
Articles
- March 2025: Gemini 2.5: Our most intelligent AI model (The Keyword)
- March 2024: New ways we’re tackling spammy, low-quality content on Search (The Keyword)
- March 2024: Google is starting to squash more spam and AI in search results (The Verge)
- September 2023: Improving Trust in AI and Online Communities with PaLM-based Moderation (Google Cloud Blog)
- July 2023: Google’s Jigsaw was trying to fight toxic speech with AI. Then the AI started talking (Fast Company)
- June 2021: Google Hopes AI Can Turn Search Into a Conversation (Wired)
- May 2021: Google isn’t ready to turn search into a conversation (The Verge)
- May 2021: Language models like GPT-3 could herald a new type of search engine (MIT Technology Review)
Blog Posts
- December 2022: Accelerating text generation with Confident Adaptive Language Modeling (CALM) (Google Research Blog)
- October 2022: UL2 20B: An Open Source Unified Language Learner (Google Research Blog)
YouTube Videos
- April 2022: Transformer Memory as a Differentiable Search Index (Machine Learning Research Paper Explained) (Yannic Kilcher YouTube Video)
- November 2021: ExT5: Towards Extreme Multi-Task Scaling for Transfer Learning (Paper Explained) (Yannic Kilcher YouTube Video)