NVIDIA Introduces Task-Seeded SDG for Nemotron Training
NVIDIA's task-seeded synthetic data generation improved multiple benchmarks by up to 11.1% in a 100B-token experiment with Nemotron-3 Nano.
Research
190 articles
In-depth coverage of new AI research — papers, benchmarks, and breakthroughs from leading labs and academia, summarized for fast reading and grounded in the methods that matter.
LATEST NEWS
-
-
Jeff Bezos Funds Flourish's Brain-Inspired AI Research
Jeff Bezos invested $50 million in Flourish, a neuro AI startup, to develop synthetic intelligence that mimics the human brain's efficiency and learning capabilities.
-
Mathematicians Warn of AI Threats to Research as Tech Industry Encroaches
A group of mathematicians issued a warning about AI's growing influence on their field, citing risks to research integrity and academic freedom.
-
Intel Explores Tabular Data Prompting for RAG LLMs
Intel researchers show tabular data improves RAG LLM performance by 12% in experiments conducted in July 2024.
-
IBM Introduces Quantum Computing Model for Complex Simulations
IBM announced a new quantum computing model capable of simulating complex physical systems with 100% accuracy in 2026, marking a significant leap in quantum algorithm development.
-
IBM, Cleveland Clinic, RIKEN Use Quantum Computing to Simulate 12,635-Atom Protein Complex
Researchers achieved a 40-fold increase in system size and 210x improvement in accuracy in quantum chemistry simulations using IBM quantum computers and supercomputers. Date: May 5, 2026.
-
IBM Quantum Heron Processor Runs 52-Qubit QFT Circuit
Researchers from ParityQC executed a 52-qubit quantum Fourier transform on IBM Quantum Heron r3, setting a new benchmark for quantum algorithm performance.
-
IBM and MIT Develop Quantum Algorithm for High-Speed Racing
IBM and MIT researchers unveiled a quantum algorithm capable of simulating high-speed racing scenarios 10 times faster than classical methods, published on April 30, 2026.
-
IBM Unveils AI-Driven Simulation Tool for Dallara
IBM introduces AI-powered simulation tool for Dallara, reducing crash-test time by 70% in early trials.
-
Intel Explains Retrieval Augmented Generation
Intel outlines how retrieval augmented generation improves AI accuracy by integrating external data sources.
-
Intel Shares Four Data Cleaning Techniques for LLM Performance
Intel outlines four data cleaning methods that improved LLM performance by up to 15% in internal testing.
-
MiniMax M3 Uses Sparse Attention for 9.7× Prefill Speedup
MiniMax's M3 model achieves 9.7× prefill speedup and 15.6× decode speedup at 1M tokens, signaling a shift toward sparse attention mechanisms.