scalable-training

Terminal-Universe is a framework that reconstructs executable workspaces from agent trajectories to synthesize diverse training tasks, improving AI agent post-training performance through supervised fine-tuning.

IMPACT is a scalable framework for training interaction-aware world models by using cross-attention as an internal interaction map to reweight denoising supervision, improving performance without external representations or inference-time changes.

SolarWM introduces an open framework and unified training recipe for building interactive video world models with scalable training across diverse data sources, enabling long-horizon real-time rollouts.

A new research paper introduces AgentMercury, a scalable framework for synthesizing executable environments from business scenarios, which improves agent training performance on various benchmarks.

FACET is a framework for synthesizing high-quality terminal tasks for AI agent training by preserving source intent and ensuring cross-artifact consistency, leading to improved performance on Terminal-Bench 2.1.

Proposes ACE, a plug-and-play method that adaptively enhances coarsening-based GNN training on heterophilic graphs by reconstructing node features and applying anisotropic regularization, achieving consistent gains on heterophilic benchmarks with minimal overhead.

LiteResearcher is a scalable reinforcement learning training framework designed for deep research agents.

Prime Intellect released prime-rl v0.6.0, enabling reinforcement learning at trillion-parameter MoE scale with sub-5-minute step times and optimized inference, training, and rollout.

An open-source repository called train-llm-from-scratch enables training billion-parameter LLMs on a single GPU, with a configurable pipeline from raw text to inference, including dataset streaming and checkpointing, under MIT License.