How OXR's RSI loop auto-labels its own failures to retrain without humans
JaynitMakwana · x · 2026-10-23
OXR claims to be the first document parsing model trained with Recursive Self-Improvement (RSI) in post-training. The loop: run multiple rollouts per page and flag wrong or disagreeing results; have a general VLM explain what went wrong; cluster explanations into recurring problem types; generate new image-label pairs per problem styled after real samples; retrain and repeat. Each round targets exactly what the model got wrong, with no hand-labeled pages.
Related event: OXR: Document Parsing Model with Recursive Self-Improvement(2 posts)→
More from Models
- Why all the AI model pwnings this week? Hype-chasing plus expiring disclosure windows — herbiebradley · 2026-10-23
- Rumor: Kimi K3.1 imminent, tipped to excel at 3D rendering — bindureddy · 2026-10-23
- Pleias wins Nvidia codefest with 2B medical models offering 10x shorter reasoning — JFPuget · 2026-10-23
- Kimi K3.1 rumored to drop soon: Opus-class agentic coding at much lower cost — bindureddy · 2026-10-23
- Ohio State's guide exposes the flaw in Microsoft Copilot Cowork's prepaid credit model — YvesMulkers · 2026-10-23
- Antigravity IDE spotted with Gemini 4 Argon and Claude Opus/Sonnet 5.5 — weskambale · 2026-10-23