Official platform of Tsinghua Big Data Research Center, sharing the team's latest research, teaching updates, and big data news.
Meta has imposed internal limits on engineers' use of external AI coding assistants Claude Code and Codex because the company fears that outputs from these tools could be unintentionally fed into its own model training and evaluation pipelines, raising legal and competitive‑risk concerns.
This survey argues that modern LLM agents should be viewed as a coupled system of a foundational model and an execution harness, analyzes the evolution from prompt engineering to harness engineering, defines six core harness responsibilities, examines task pressures, proposes richer evaluation metrics, and outlines future research directions.
The paper introduces Counterfactual Policy Optimization (CFPO), a training‑time framework that inserts causal consistency constraints into multimodal reinforcement learning, forcing vision‑language models to rely on essential visual evidence and achieving consistent accuracy gains across real‑world and math‑centric benchmarks.
Polars outperforms Pandas on large‑scale ETL by using a multithreaded lazy execution model, columnar Arrow storage, and query optimizations, delivering up to 94× speedups on 10 GB workloads, while Pandas remains suitable for smaller datasets and tight ML ecosystem integration.
This article surveys the four major AI engineering paradigms—Prompt, Context, Harness, and Loop—detailing their technical logic, practical implementations, trade‑offs, and real‑world incidents, while providing concrete guidelines and comparative analysis for building autonomous AI systems.
On June 30 2026 Anthropic introduced Claude Science, an AI‑powered workbench that unifies common research tools, generates auditable outputs, offers built‑in genomics, proteomics and cheminformatics skills, and manages scalable compute on HPC or cloud, enabling scientists to conduct end‑to‑end analyses from data acquisition to manuscript preparation.
The article explains why large language models need a physical embodiment to move beyond text, outlines the three core components of embodied AI—multimodal brain, sensor fusion, and actuators—reviews recent breakthroughs such as Google RT‑2 and Sim2Real, and explores how these systems could transform homes, factories, and extreme environments.
The article announces the winning teams of the 2026 China University Computer Competition Big Data Challenge and provides detailed, step‑by‑step experience reports covering data processing, feature engineering, model design, training strategies, and post‑processing for a cross‑sectional stock ranking task.
The 2026 China University Computer Competition's Big Data Challenge revealed the monthly star award winners, each receiving a prize, and the top teams detailed their feature engineering, model selection, training validation, and portfolio combination tactics that led to their success.
The article surveys the leading embodied‑intelligence datasets—including ALOHA, Open X‑Embodiment, AGIBOT WORLD, RH20T and RoboMIND—detailing their technical features, scale, multimodal content, and how they support model training, benchmark evaluation, and practical robot development pipelines.
