TheoremSmith: Automating Hard Proof Generation to Scale LLM Formal Reasoning
Charles Huang's TheoremSmith system automates the creation of formally verified mathematical proof problems to train language models on symbolic reasoning at scale.
Charles Huang's TheoremSmith system automates the creation of formally verified mathematical proof problems to train language models on symbolic reasoning at scale.
Nemotron-Personas-Finance-Japan pairs official statistics with household finance data to enable privacy-safe AI development for financial services.
Physicl, founded by a former e-commerce 3D rendering executive, secured seed funding to help roboticists generate high-fidelity synthetic training data through its Studio platform.
NVIDIA open-sources specialized variants of Cosmos 3 Nano for autonomous driving and robotics, enabling multi-sensor synthetic data generation and physics-based video prediction.
IBM Research demonstrates that a 20B-parameter model fine-tuned on synthetic data can generate on-brand slide decks while remaining deployable on private infrastructure.
A joint effort brings curated synthetic pretraining data across 22 Indian languages to open-source, addressing a critical gap in non-English language model training.