AI
RoboFoundry: System-as-Policy Evolution for Self-Learning Embodied Agents
Publication: Hugging Face Daily Papers
Author: Not available in collected metadata
Source published: 25/09/2026, 20:00:00 UTC
Briefing created: 29/09/2026, 03:26:25 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 29/09/2026, 03:18:24 UTC.
Briefings
RoboFoundry is a new framework that treats the entire agent system as a policy, using execution traces to automatically update system capabilities rather than optimizing isolated components. On EmbodiedBench, it improves GPT-5.5 performance by 27.8% and brings Qwen3.7-Plus to near parity with GPT-5.5. On LIBERO-PRO, it outperforms Cap-Agent0 by 243.8% to 679.7% across various perturbation types. The framework also significantly enhances long-horizon memory, outperforming baselines by at least 39.0% on RoboMemArena. It enables zero-shot transfer and online evolution across different robots and tasks.
Key points
- RoboFoundry diagnoses capability gaps in decision-making and memory management, converting execution traces into validated system updates.
- On EmbodiedBench, RoboFoundry improves GPT-5.5 performance by 27.8% and brings Qwen3.7-Plus to near parity with GPT-5.5 (70.3% vs. 72.7%).
Context and limitations
Performance gains are measured on specific benchmarks (EmbodiedBench, LIBERO-PRO, RoboMemArena) and may not generalize to all real-world deployment scenarios or heterogeneous robot setups.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 29/09/2026, 03:26:25 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeibay2xymfv2ycw65a7yavupcle35qsxibwkzm233ung6ndozyxvzq