Synthetic data generation

Technique · Training · 2 items since Jun 2026

Training on data generated by models, such as rewritten captions or simulated agent traces.

Trends

Jul 24
Apple synthesizes API-agent data without environments
An LLM acting as an on-the-fly world model synthesizes agent trajectories from API specifications alone, yielding fine-tuning gains on AppWorld and OfficeBench without any executable backend.
Jun 8
Microsoft Research's Lens image model
A 3.8B text-to-image model shows detailed training captions can substitute for raw scale, matching or beating much larger models on several benchmarks with roughly one-fifth the pre-training compute; code and checkpoints are MIT-licensed.