Tencent Backs Alibaba's Former Qwen Researcher in $20M AI Lab Deal

Researchers from UC Berkeley, Princeton, EPFL, and Databricks introduced PixelRAG, a retrieval system that bypasses traditional text parsing by rendering web pages as screenshots and indexing them directly for vision-language models. Tested on 30 million Wikipedia screenshot tiles, PixelRAG improved accuracy by up to 18.1% over text-based RAG systems and reduced token costs by 10x. The approach addresses fundamental information loss in conventional HTML-to-text conversion pipelines.

Google researchers propose 'faithful uncertainty,' a technique that allows large language models to express qualified guesses rather than either confidently hallucinating or refusing to answer. The approach reframes hallucinations as 'confident errors' and enables models to hedge responses appropriately, preserving utility while maintaining trustworthiness. This addresses a core tradeoff in LLM deployment where eliminating factual errors typically forces models to abstain from answering questions they actually know.

Yen-Ling Kuo, an assistant professor at the University of Virginia, received the IEEE Robotics and Automation Society's inaugural Outstanding Women in Robotics and Automation Early Career Contribution Award for her work on uncertainty estimation in robotic manipulation. Her research method, detailed in the paper 'Diff-DAgger: Uncertainty Estimation with Diffusion Policy for Robotic Manipulation,' enables robots to make informed decisions in unfamiliar scenarios while reducing the need for human supervision. The approach improves task completion rates and creates pathways for more complex models in interactive robot learning.

Researchers from NYU, Columbia, Princeton, University of Maryland, Harvard, and Lawrence Livermore National Laboratory published a paper introducing Latent Context Language Models (LCLMs), a compression technique that reduces LLM input by 16x while maintaining accuracy better than existing methods. Unlike KV cache compression, LCLMs compress tokens before decoder processing, delivering 8.8x faster output on long-context benchmarks. The models are open-sourced on HuggingFace and designed to integrate into existing LLM stacks.