机器之心 JIQIZHIXIN

机器之心 JIQIZHIXIN

Users
Tweets

机器之心 JIQIZHIXIN

@jiqizhixin

Is your AI model stuck because of bad data, not bad algorithms? Researchers from Shanghai Jiao Tong University, Carnegie Mellon University, and other top institutions present DataMaster: an autonomous agent that treats data engineering like a tree search—branching to discover external datasets, selecting and cleaning them, then reusing what works via shared memory and a central pool. Results: On MLE-Bench Lite, it boosts the medal rate by 32.27% over the starting score. On PostTrainBench, it edges past the instruct model on GPQA (31.02% vs 30.35%). No algorithm changes needed. DataMaster: Data-Centric Autonomous AI Research Project: github.com/sjtu-sai-agents/D… Paper: arxiv.org/abs/2605.10906 Our report: mp.weixin.qq.com/s/V4eNAG8v2… 📬 #PapersAccepted by Jiqizhixin

1,838

Wagner Meters

Wagner Meters @WagnerMeters

Jun 10

The DataMaster L6 is a free mobile app that stores moisture data and creates moisture reports so you can protect your data and predict moisture trends.🤩 Get started today by downloading it: wagnermeters.com/datamaster/

0:47

Yaxin Du

Yaxin Du @DorothyDDU

May 18

Excited to share our new paper: DataMaster — an AI data engineer for machine learning. The key question is simple: Can AI improve a model without changing the model architecture or training algorithm, but only by improving the data? Paper: arxiv.org/abs/2605.10906

Machine Learning (ML) Papers

Machine Learning (ML) Papers @Memoirs

May 14

DataMaster: Towards Autonomous Data Engineering for Machine Learning Yaxin Du, Xiyuan Yang, Zhifan Zhou, Wanxu Liu, Zixing Lei, Zimeng Chen, Fenyi Liu, Haotian Wu, Yuzhu Cai, Zexi Liu, Xinyu Zhu, WenHao Wang, Linfeng Zhang, … arxiv.org/abs/2605.10906 [𝚌𝚜.𝙻𝙶 𝚌𝚜.𝙰𝙸]

As model families, training recipes, and compute budgets become increasingly standardized, further gains in machine learning systems depend increasingly on data. Yet data engineering remains largely manual and ad hoc: practitioners repeatedly search for external datasets, adapt them to existing pipelines, validate candidate data through downstream training, and carry forward lessons from prior attempts. We study task-conditioned autonomous data engineering, where an autonomous agent improves a fixed learning algorithm by optimizing only the data side, including external data discovery, data selection and composition, cleaning and transformation. The goal is to obtain a stronger downstream solution while leaving the learning algorithm unchanged. To address the open-ended search space, branch-dependent refinement, and delayed validation inherent in autonomous data engineering, we propose DataMaster, a data-agent framework that integrates tree-structured search, shared candidate data, and

ALT As model families, training recipes, and compute budgets become increasingly standardized, further gains in machine learning systems depend increasingly on data. Yet data engineering remains largely manual and ad hoc: practitioners repeatedly search for external datasets, adapt them to existing pipelines, validate candidate data through downstream training, and carry forward lessons from prior attempts. We study task-conditioned autonomous data engineering, where an autonomous agent improves a fixed learning algorithm by optimizing only the data side, including external data discovery, data selection and composition, cleaning and transformation. The goal is to obtain a stronger downstream solution while leaving the learning algorithm unchanged. To address the open-ended search space, branch-dependent refinement, and delayed validation inherent in autonomous data engineering, we propose DataMaster, a data-agent framework that integrates tree-structured search, shared candidate data, and

188

#Lula com #Chuchu

#Lula com #Chuchu @fferreiralimma

Apr 11

Replying to @pesquisas_elige @FlavioBolsonaro @LulaOficial

Piada pronta do DataMaster LULA REELEITO COM CERTEZA ABSOLUTA SOU LULA E HADDAD

201

DaimonRobotics

DaimonRobotics

@DaimonRobotics

Apr 9

🎬Unboxing time - DataMaster Pack (Controlpad Version) ✨A plug-and-play teleoperation data acquisition system — simple, efficient, and built for real-world robotics. #DaimonRobotics #Robotics #Teleoperation #EmbodiedAI #DataCollection

2:20

242

Huawei IT Products & Solutions

Huawei IT Products & Solutions

@IT_Huawei

Mar 3

x.com/i/article/202864109498…

10,588

AI Vanguard

AI Vanguard @YsJaganTweets

Jan 23

I am nominating .@BugganaRaja as DataMaster✊to Expose reality of Amaravati & Debts📈✊& No 6 Sir hold pressmeets in Amaravati & compare Amaravati budget allocations vs past development activities ⭕ Expose every debt with unique pressmeet style🙏

Graduate Adda

1:30

AI Vanguard @YsJaganTweets

Jan 22

I am nominating .@gudivadaamar as Expose of PR govt📈 with PR& No 5 kutami govt trolled & misleaded ur statements But u give return gift with video pressmeets before every industries under our tenure 19-24 & future mous⭕ Just go & expose lokesh💯

1:26

374

8,379

baiwei

baiwei @baiweiart

Jan 19

Provenance of datamaster @Vantekai_ ⭕️

baiwei @baiweiart

4 Oct 2025

GM! 🌞 On provenance seals ⭕️ and blockchain ⛓️ In ancient Chinese art signatures authenticate and ensure artwork’s ownership tracking provenance over it. Blockchain echoes this tradition, allowing expansion to network art. 「夜白圖」~750AD —「ZATO Provenance」2024

121

CrimeWatchMpls

CrimeWatchMpls

@CrimeWatchMpls

11 Oct 2025

DWI cases called into question after testing machines across the state discovered to have data entry errors. BCA orders immediate statewide inspection of DWI testing instruments, updates procedures to prevent future errors ST. PAUL — The Minnesota Department of Public Safety’s Bureau of Criminal Apprehension (BCA) is requiring local law enforcement agencies across Minnesota to immediately review and verify dry gas cylinder data has been entered correctly into DataMaster (DMT) instruments currently being used. This requirement follows the discovery of multiple data entry errors by trained operators during the installation of gas cylinders in the instrument. As part of this review, BCA procedures for maintaining DMT instruments will also be updated to prevent future errors. Out of an abundance of caution, the BCA has ordered the suspension of all DMT usage until agencies verify the gas cylinder data is correctly entered into each instrument. Recognizing the potential of human error with the existing exchange process, all gas cylinder replacements and maintenance moving forward will be conducted solely by BCA personnel, effective immediately. “The BCA is committed to the highest level of quality in our forensics lab, and this is why we are conducting this review,” said BCA Superintendent Drew Evans. “We appreciate our law enforcement partners’ prompt attention to this inspection process.” Select Minnesota law enforcement personnel are trained to administer breath tests and maintain the DMT instrument as part of impaired driving enforcement. Each DMT instrument holds a uniquely numbered dry gas cylinder with a known alcohol concentration, which serves as a control during every test to ensure the instrument is measuring alcohol accurately. When the instrument is first set up, authorized law enforcement operators enter information from the cylinder into the system. That information remains in use for all subsequent tests until the cylinder is replaced, at which point new cylinder information must be entered. It was recently discovered that, in some cases, incorrect cylinder information had been entered into the instrument, resulting in test data being called into question. The BCA became aware on Friday of two more instruments with data entry errors related to gas cylinder installation. These instruments are in Olmsted County, where the error may affect 108 cases, and Hennepin County, where the error may affect 38 cases. This follows the discovery of errors in Aitkin, Winona and Chippewa counties. Agencies that house the instruments are not necessarily responsible for the error as multiple law enforcement agencies use each instrument. The BCA is dedicated to partnering with Minnesota law enforcement agencies to keep Minnesota roads safe from impaired drivers. As part of that effort, the BCA will continue providing the training and information necessary to ensure the accuracy and reliability of all breath testing in Minnesota.

127

19,494