AI Briefing#synthetic-data
The latest AI and developer news about #synthetic-data, with the original source and a short summary.
The latest AI and developer news about #synthetic-data, with the original source and a short summary.
Mitra-v2, a second-generation tabular foundation model, is pretrained on tables with 10x more rows and 3x more columns than its predecessor, leading TabArena and ScoringBench.
The CC BY 4.0 licensed dataset contains 50 fields per record and is built using NeMo SDG-PGMs and gemma-4-31B-it to support privacy-compliant financial AI training.
The platform uses publicly available data sources like O*NET and SEC filings to generate realistic test scenarios without using customer data.
The expanded platform includes five cloud-native data blueprints, support for fine-tuning open models like GR00T and QWEN, and Cosmos 3 Nano variants that enable up to 2x faster inference.
Fine-tuned on 2B tokens of synthetic data, the model predicts screen deltas from keypresses, outperforming prompted baselines on shell and file readback tasks.
The NeurIPS 2026 paper provides a library of synthetic time series generators and augmentations for training a DeltaProduct based linear RNN.
The framework achieves a KL divergence of 1.01 against real user queries, a 9.2x improvement over previous baselines.
A 300M-parameter transformer trained on synthetic recurrent sequences achieves 0.9–2.4 bits per byte on real languages after reading 1 million bytes.
The framework improves Pass@1 scores by up to 8.4 percentage points in ITSM and Hybrid enterprise environments by targeting model capability gaps.
LG AI Research has released EXAONE Tabular, a lightweight tabular foundation model pre-trained without real data.
LG AI Research has released 'EXAONE Tabular', a foundation model for tabular data pre-trained exclusively on synthetic data without using any real-world data.
Apple has released Agent Seer, a tool that automatically generates AI agent evaluation scenarios without manual effort, using only MCP specifications.
LangChain has released a two-stage pipeline that automatically generates environments and tasks for agent evaluation.
A search DL model was completed in 3 weeks using AI agents and automation loops.
NVIDIA supports Physical AI development and validation with open world models and Omniverse.
Amazon is providing Trainium research support to 34 AI research teams from 30 universities.
F.R.A.N.T.Z. extracts scene semantics from monocular video without synthetic data or task-specific training.
This summarizes the role of simulation and its technical paradigms as key to securing data, which is central to Physical AI development.
LG AI Research proposed an OOD detection methodology that generates sophisticated synthetic outliers using a Diffusion model.
Alibaba Tongyi Lab's WebDancer proposes a methodology that enhances LRMs' long-horizon reasoning capabilities by leveraging web search tools.
StradVision built a synthetic data generation pipeline for rare scenarios by leveraging AWS cloud and a hybrid architecture.
NVIDIA has released the Nemotron open datasets and a visualization tool to improve agent performance.
General Intuition, Kyutai, and Epic Games collaborated to release MIRA, a 5-billion-parameter multiplayer world model.
This introduces research that improves HIP kernel generation performance for AMD GPUs using synthetic data and reinforcement learning.