Tagged articles
Data Synthesis
26 articles · Page 1 of 1
Firefly introduces a back-chaining pipeline that explores real MCP servers to generate verifiable tool-call tasks, builds a retrieval-augmented simulator for reproducible offline RL, and shows a 4B model reaching Claude-level performance on tool-use benchmarks.
Back-ChainingData SynthesisMCP Servers
0 likes · 27 min read

Shanghai Jiao Tong University researchers introduce ZwZ2 and the EvidenceConnect framework, which synthesizes 20.4K single-image training samples requiring models to connect evidence across multiple regions, yielding 4B, 9B, and 27B models that surpass base models on 25 benchmarks and transfer gains to multi-image, spatial reasoning, and agent tasks.
Data SynthesisEvidenceConnectGRPO
0 likes · 14 min read

Researchers from USTC and Alibaba DreamX introduce TerminalHorizon, an automated data synthesis engine that generates 2,852 long-horizon training trajectories across 25 domains and 27 behavioral patterns, enabling Qwen3.5-27B to achieve 60.67% on Terminal-Bench 2.1 and sustain 12-hour autonomous research in RSI exams.
Data SynthesisQwen3.5-27BRSI exam
0 likes · 17 min read

This article analyzes four test data generation paradigms—rule-driven, model-driven, AI-enhanced, and contract-collaborative—with real-world case studies from banking, e-commerce, insurance, and ride-hailing, revealing key decision factors for technology selection and future trends like AI copilots and standardized quality metrics.
AI-EnhancedContract-DrivenData Synthesis
0 likes · 8 min read

The article explains how the DOMINO method lets large language models learn a domain from just dozens of real examples instead of hand‑written prompts, describes its trainable "domain switch" architecture, and shows experimental gains on time‑varying code tasks, highlighting more robust and diverse data synthesis.
Data SynthesisDominoKDD2026
0 likes · 8 min read

The article outlines a practical pipeline for creating agentic factual SFT and mid‑train datasets, covering how to define training goals, filter and classify queries, label processing tags, format trajectory samples, differentiate SFT from mid‑train data, and avoid common pitfalls when generating evidence‑driven AI training data.
Agentic AIData SynthesisSFT
0 likes · 10 min read

This article outlines a systematic approach for creating agentic factual SFT and Mid‑train data, covering the definition of training goals, query filtering, two‑layer classification and labeling, trajectory format, differences between Mid‑train and SFT, a practical synthesis pipeline, and common pitfalls to avoid.
Agentic AIData SynthesisSFT
0 likes · 11 min read

The MMKG‑RDS framework introduced by 360 AI Lab creates a complete pipeline—from multimodal document parsing and knowledge‑graph construction to customizable task synthesis and multi‑dimensional quality assessment—enabling the production of high‑quality reasoning data that significantly boosts large‑model performance across diverse domains.
AI ReasoningData Synthesisknowledge graph
0 likes · 7 min read

EvoCUA, a native computer‑use agent from Meituan, combines a verifiable data‑synthesis engine, a ten‑thousand‑level sandbox infrastructure, and an experience‑driven learning paradigm to overcome data‑scaling and feedback challenges, achieving a 56.7% success rate on the OSWorld benchmark and surpassing previous open‑source models.
AI agentData SynthesisOSWorld
0 likes · 27 min read

The paper presents ShoppingBench, a large‑scale e‑commerce benchmark built on a 2.5 million‑product virtual sandbox, evaluates 17 LLM agents (with GPT‑4.1 achieving under 50 % success), and introduces ORM‑Virtual environment, scalable data synthesis, trajectory distillation and RL to train a lightweight model that rivals GPT‑4.1, now deployed in LazzieChat.
Agentic AIData SynthesisLLM
0 likes · 12 min read

The paper introduces TradingGroup, a five‑agent LLM‑based quantitative trading framework that incorporates a self‑reflection mechanism, dynamic risk management, and an automated data‑synthesis pipeline, and demonstrates superior cumulative returns, Sharpe ratios, and lower drawdowns than rule‑based, ML, RL, and existing LLM strategies on five real‑world stock datasets.
Data SynthesisLLMfinancial AI
0 likes · 14 min read

In an interview before the DACon conference, Dr. Feng Ziyong reveals how Glint‑MVT and novel data‑synthesis techniques overcome distribution gaps, improve compositional understanding, and enable billion‑scale, second‑level retrieval for city‑level surveillance, while balancing model efficiency and effectiveness.
Data Synthesiscity surveillanceembedding retrieval
0 likes · 11 min read

This article explores how high‑quality inference data has become a new paradigm driving AI breakthroughs, detailing Ant Group's research on inference data paradigms, financial‑sector applications, intelligent labeling and quality inspection, and the AIGD AI data synthesis platform, followed by a technical Q&A.
AI dataAIGDData Synthesis
0 likes · 11 min read

This article surveys recent large‑scale corpus rewriting techniques for LLM pre‑training, covering K2’s token‑utilization strategies, domain‑specific methods like SwallowMath/Code, reStructured pretraining, the WRAP pipeline, Nemotron‑CC filtering, Pro‑X noise removal, and the MAGA multi‑style expansion, while highlighting challenges, experimental findings, and open research questions.
Data SynthesisLLMPretraining
0 likes · 20 min read

The article introduces Easy Dataset, a GUI‑driven framework that transforms heterogeneous documents into high‑quality, persona‑driven fine‑tuning data for large language models, details its architecture, core contributions, experimental validation on financial QA, and compares it with existing data‑synthesis tools.
Data SynthesisGUILLM
0 likes · 12 min read

This work tackles the difficulty of incorporating extensive domain knowledge into in‑domain NL2SQL tasks by proposing an intermediate‑representation‑based data synthesis method that decouples knowledge compliance from SQL generation, enabling automated creation of high‑quality training data with 60× human efficiency and over 97% accuracy.
Data SynthesisNL2SQLSQL Generation
0 likes · 2 min read


DevOpsMay 18, 2025 · Artificial Intelligence
Although large language models have enabled AI agents that mimic human digital interactions, their commercial accuracy remains far below production standards, prompting the industry to pivot toward agentic workflows and data synthesis, which promise more reliable task automation, reasoning, and observable, auditable processes for knowledge work.
Agentic workflowsData Synthesisknowledge work
0 likes · 6 min read

The article explains that low accuracy and security risks of current AI agents—evidenced by a Claude AI Agent achieving only 14% of human performance and an average success rate of about 20%—are driving a move toward agentic workflows, which offer observable, auditable, and data‑synthesizing pipelines that dramatically improve enterprise productivity.
AI agentsAgentic workflowsData Synthesis
0 likes · 7 min read

This article presents the ORLM intelligent decision‑making large model, detailing how real‑world decision problems are formalized and solved, the training difficulties and data synthesis methods, the transition from academic research to commercial platforms, and future technical improvement plans.
AIData SynthesisModel Training
0 likes · 10 min read

This article surveys the post‑training pipelines of major open‑source large language models released this year, detailing their alignment algorithms, data synthesis, reward modeling, DPO/GRPO variants, long‑context handling, tool use, and model‑averaging techniques, and highlights emerging trends such as data‑centric pipelines and iterative weak‑to‑strong alignment.
AI researchData SynthesisLLM
0 likes · 99 min read

This paper introduces Parrot, a system that enhances large language models' (LLMs) multi-turn instruction following capabilities through context-aware preference optimization (CaPO) and synthetic data generation, achieving significant performance improvements with limited training data.
CaPOData SynthesisNLP
0 likes · 9 min read

This article explores how generative AI (AIGC) can be used to synthesize structured data, how such synthetic data enhances causal inference tasks, and how agent‑based modeling and the YLearn framework together enable a two‑way synergy between AIGC and causal learning for enterprise AI solutions.
AIGCData SynthesisYLearn
0 likes · 15 min read

ShopeeVideo OCR is a multi‑language text‑recognition system for Southeast Asian e‑commerce videos that unifies detection, Transformer‑based recognition, layout analysis, and large‑scale synthetic data generation to handle Indonesian, Filipino, English, Vietnamese, Thai and Chinese scripts, delivering industry‑leading accuracy and winning thirteen ICDAR first‑place awards.
Computer VisionData SynthesisMulti-language OCR
0 likes · 15 min read

This article describes a comprehensive solution for detecting checkboxes in document images and determining their selected or unselected status by combining YOLOv5 object detection, synthetic and semi‑synthetic data generation, specialized post‑processing, and association logic to handle varied shapes, positions, and markings.
Data SynthesisObject DetectionYOLOv5
0 likes · 13 min read

Ant’s self‑developed xNN‑OCR demonstrates how advanced OCR can run offline on smartphones by combining GAN‑based data synthesis, lightweight ShuffleNet‑inspired detection, NAS‑optimized recognition, and aggressive model compression, delivering near‑real‑time accuracy for diverse mobile scenarios while preserving privacy and low cost.
Data SynthesisNASedge AI
0 likes · 11 min read

The article summarizes Andrew Ng’s presentation at BADLS, highlighting major deep‑learning trends such as the rise of big data, end‑to‑end models, the bias‑variance tradeoff, human‑level performance benchmarks, and practical advice for improving one’s AI skills.
AI TrendsData SynthesisEnd-to-End
0 likes · 10 min read
