Official platform of Tsinghua Big Data Research Center, sharing the team's latest research, teaching updates, and big data news.
A recent study shows that AI not only consumes energy in data centers but also amplifies fossil‑fuel productivity, leading to an estimated 0.47 billion‑ton rise in global CO₂ emissions, highlighting the need to monitor AI’s broader impact on the energy sector.
This survey examines proactive protection methods—ranging from privacy filters and non‑learnable samples to generative safeguards, adversarial captchas, and traceability mechanisms—that embed adversarial perturbations before visual content is shared, trained, generated, accessed, or disputed, and evaluates them across transferability, adaptability, and deployment maturity.
Google DeepMind’s WeatherNext AI model, trained on 20 TB of global atmospheric data and a curated historical cyclone database, can predict tropical cyclone tracks, intensity and structure up to a day earlier than traditional systems, generating ensemble forecasts up to 15 days ahead with competitive accuracy.
The RL‑100 framework combines imitation learning from human tele‑operation with offline and online reinforcement learning to refine diffusion‑based control policies, achieving 100 % success across eight real‑world robot tasks, matching or surpassing human operators in speed while maintaining stability and low latency.
This survey examines how large‑language‑model‑driven AI scientists now span the full research lifecycle, yet most systems provide scant evidence for reproducibility and claim verification, analyzing 35 works to reveal audit gaps and propose a concrete reporting checklist for trustworthy autonomous research.
This survey maps a decade of robot‑learning research onto a weight‑vs‑skill axis, classifies 77 representative systems into six technical branches, analyzes their trade‑offs, highlights emerging skill‑economy challenges, and proposes measurable metrics for future self‑improving robotic systems.
This article reviews a PhD thesis that systematically studies online and offline learning for contextual bandits with huge action spaces, highlighting statistical, computational, and optimization challenges and presenting mixed‑effect Thompson sampling, diffusion priors, structured direct methods, and PAC‑Bayes pessimism as effective solutions.
This article reviews the paper “Causal Inference with Unstructured Outcomes”, explaining how the authors extend causal analysis from scalar results to text and image data by defining a max‑contrast feature, presenting identification conditions, estimation algorithms, and extensive experiments on formalness, toxicity, image blur, and paired treatment‑result scenarios.
The article explains a three‑layer Skill architecture for AI agents, shows how real user feedback is turned into concrete rule updates across routing, instruction, and resource layers, and describes iterative refinement, compaction, and validation before releasing new Skill versions.
The article explains how to construct a simple radio telescope using household materials, capture the 21‑cm hydrogen line, analyze Doppler‑shifted spectra to measure cloud velocities at various galactic radii, and demonstrate the flat rotation curve that signals the presence of dark matter.
