Plurai raises $13 million to simulate and tune AI agents
The seed round, led by Team8, funds digital twins that post-train small open-weight models for customer workflows.
Stories per day · 30 days
5 RSS
Relateddifferential-privacy5reinforcement-learning4benchmark2grpo2long-context23d1agents1ai-agents1benchmarks1cancer-research1chebyshev-moments1clinical-notes1data-agents1dataset1dataworld1de-identification1diffusion1distillation1
55 stories, newest first · one card per story
Stories tagged #synthetic-data
Nemotron-Personas-Finance-Japan synthetic dataset covers Japan
The Hugging Face post, under an NVIDIA path, gives no further details beyond the title.
Audience-conditioned privatization for private text generation
Ext-Classify-Priv framework classifies attributes by audience to preserve only contextually sensitive information in differentially private synthetic data releases.
arXiv presents biology-informed framework for evaluating synthetic ovarian cancer multi-omics datasets
The framework evaluates paired RNA expression and CNV profiles from the DECIDER cohort using variational autoencoders.
arXiv releases CoRTeC for differentially private synthetic tabular data
CoRTeC releases synthetic data under differential privacy using noisy counts, decodable without language models or real records.
SPLICE: Differentially Private Synthetic Data with Dimension-Independent Guarantees
SPLICE returns a private distribution with Sliced Wasserstein-1 distance tilde O(1/sqrt(n)) from the target, independent of data dimension d.
A-RSI refreshes tasks for zero-data multimodal self-improvement
On an eleven-benchmark suite the authors report the average rising from 57.77 to 61.05 using 800 fresh tasks.
SynthAPI scaling law links tool-use gains to environment diversity
Across 458,752 tasks and 849 models, tool and database diversity beat adding tasks as data budgets grow.
PRISMA builds differentially private synthetic in-car energy traces
At epsilon 0.5, PRISMA beats tabular DP baselines by keeping diverse power profiles instead of collapsing to few modes.
EdgeDeID de-identifies clinical notes locally without real patient training data
On a 100-note synthetic benchmark it reaches 0.994 micro-F1, but F1 falls to 0.48 on MIMIC ICU nursing notes.
PHYRD benchmark tests LLMs on reproducing and extending physics
Across five models, Reproduce accuracy was 49.21%–68.25% and closed-book Discover was 18.60%–74.42%.
On-policy self-training improves models without curated answers
From coarse topic descriptions alone, a model writes its own questions and reasoning traces and gains on language and visual tasks.
NVIDIA adds Cosmos tools to train and deploy world models
New blueprints, Cosmos 3 Nano simulation variants, and NIM microservices claim up to 2x faster inference.
2
LongZero improves LLM long-context reasoning without external data
The method self-writes multi-hop and retrieval tasks, trains with GRPO, and beat external-corpus RL on five benchmarks.
LongZero improves LLM long-context reasoning without external training data
Models write compact evidence, expand it into multi-document context, and train with GRPO on self-made tasks.
TextRIRO trains scene text editing on real self-paired images
Shuffle-and-crop self-pairs train a diffusion editor, and derived TextRIRO-3M lifts long-text recognition.
Ask-to-Distill learns to write questions for on-policy distillation
A frozen student–teacher pair scores synthetic questions without reference answers, then those prompts feed standard on-policy distillation.
InfiniDream generates editable photorealistic 3D worlds with VLMs
Evaluations on 100 object and 20 scene tasks report gains on four perceptual metrics.
VerifierBench tests whether agents can write executable solution verifiers
Across 19 models, normalized balanced accuracy ranged from 10.6 to 62.1 on 222 tasks.
Study compares privacy and capability of synthetic medical-record training
Direct finetuning leaked most; DP-SGD cut leakage but hurt capability; privacy-aware rewriting balanced both better.
Data World: Scaling Data Environments from Scratch
Data World synthesizes data-agent tasks and trains DataWorld-35B-A3B, which beats its Qwen3.6 base on several benchmarks.
Scaling Agentic Data for Long-Horizon Terminal Intelligence
Researchers fine-tune Qwen3.5-27B into TerminalHorizon-27B using 1.5K long-horizon terminal environments.
SheetSage2: Coherent Lead-Sheet Transcription with Synthetic Supervision
SheetSage2 transcribes coherent lead sheets using synthetic MIDI supervision and beats prior systems on most reported metrics.
FALCON: A Model and Dataset Agnostic Framework for Synthetic Data Generation for NL2SQL Pairs
FALCON generates realistic, ambiguity-aware NL2SQL training pairs with compact open models, matching or exceeding challenging benchmark complexity locally.