InfraRSI: Recursive Self-Improvement for Optimizing Large Model Training Infrastructure
Start here
On November 9, 2026, Hugging Face Daily Papers included the paper InfraRSI. The research proposes an agent framework that transforms human instructions into verified distributed large model training optimizations and uses a fixed task contract to prevent reward hacking. On four dense and MoE workloads, it achieves up to 2.75x speedup, raising average MFU from 17.4% to 30.0%, outperforming existing tuners and coding agents. The framework forms a two-level recursive self-improvement loop by online correction of priors and cross-cluster distillation of experience.
Generated by AI from reports · Updated 2 days ago
Attention and developments
Attention, hourly
Now 25, -1% in 24 hours·Peak 27 (Nov 10 23:00)
Related research
This event's paper
Progress timeline
- Hugging Face Daily PapersInfraRSI: Optimizing Training Infrastructure with Recursive Self-Improvement
InfraRSI is an agent framework that turns human instructions into verified distributed LLM training optimizations and uses a fixed task contract to prevent reward hacking. On four dense and MoE workloads, it achieves up to 2.75x speedup, raising average MFU from 17.4% to 30.0%, outperforming existing tuners and coding agents. It corrects priors online and distills experience across clusters, forming a two-level recursive self-improvement loop.
Related discussion
Attention trend
New sources per day
- Nov 8: 1 new sources (1 outlets, 0 accounts)
- Nov 9: 0 new sources (0 outlets, 0 accounts)
- Nov 10: 1 new sources (0 outlets, 1 accounts)
- Nov 11: 0 new sources (0 outlets, 0 accounts)
- Nov 12: 0 new sources (0 outlets, 0 accounts)
- Nov 13: 0 new sources (0 outlets, 0 accounts)