RSIAgent improves agents without weight updates, beats GPT-6 Astra on OSWorld 2.0
Roger_M_Taylor · x · 2026-09-16
Researchers introduce RSIAgent, a framework for recursive self-improvement via autonomous exploration: the agent decides what to explore, executes tasks, verifies outcomes, and consolidates stable action-condition-outcome relationships into memory for reuse — dubbed Scaling Experience.
With base model weights frozen (Kimi-K3 and GLM-5.3), it still keeps improving through acquired experience:
- OSWorld 2.0 (0808 offline): 78.98% Partial Score vs 72.60% for GPT-6 Astra
- Agents' Last Exam (Near-term): 84.82% vs 82.26% for GPT-6 Astra
The takeaway: agents can keep getting better by acquiring, verifying, and reusing their own experience without any weight updates.
Related event: RSIAgent Achieves Recursive Self-Improvement Without Training(3 posts)→
More from coding & agent
- 10 must-know topics for RAG engineer interviews: chunking, retrieval metrics, hallucination debugging — ashishllm · 2026-09-16
- Third-party Grok Bot turns YouTube lectures into exam-ready cheat sheet PDFs — tetsuoai · 2026-09-16
- Jev programming likened to MapReduce for decisions: parallel, mutually unaware queries — cocktailpeanut · 2026-09-16
- 53 MCP servers scanned: 36% graded D/F, mostly for over-permissioned scope — BrilliantSecret143 · 2026-09-16
- Claude Code's auto mode quietly uses a second safety classifier model, sparking Max-tier transparency complaints — tomekkorbak · 2026-09-16
- When an LLM plans an executable agent DAG, where do you draw the trust boundary? — Repulsive_Sugar_5252 · 2026-09-16