EdgeBench: Unveiling Scaling Laws of Learning from Real-World Environments
- Published
- Source
- arXiv
- Paper number
- 567
- Field
- LLMs / NLP
- arXiv ID
- 2607.05155
Key points
- Environment learning average performance follows a log-sigmoid scaling law, with R² = 0.998, and the pattern is stable across all 6 task families.
- It studies 134 real-world tasks, each requiring 12 or more continuous hours of execution, with large tasks that take human experts an average of 57.2 hours.
- Frontier agent learning speed roughly doubles every three months, based on models released after September 2025.
- Continuous experience is stronger than independent restarts, and accumulated experience strongly determines long-horizon performance.
- The theory derives the log-sigmoid shape as a frontier-expansion process on a latent task graph.
Paper links
External research summaries. These are not HDATF publications or measured product results.