analiza aiqbench
ASH demonstrates a scalable approach to long-horizon embodied learning by eliminating the need for hand-engineered rewards or expert annotations. The system autonomously improves by training an Inverse Dynamics Model from its own trajectories to extract supervision from unlabeled internet video, enabling sustained progression in complex game environments where standard baselines plateau. This self-improvement loop effectively leverages unsupervised learning to build long-term memory, offering a promising direction for autonomous agents tackling multi-hour planning tasks.
ASH: Agents that Self-Hone via Embodied Learning
arXiv:2605.14211v3 Announce Type: replace-cross Abstract: Long-horizon embodied tasks remain a fundamental challenge in AI, as current methods rely on hand-engineered rewards or action-labeled demonstrations, neither of which scales. We introduce ASH, an agentic system that learns an embodied policy from unlabeled, noisy internet video, without reward shaping or expert annotation. ASH follows a self-improvement loop; when it gets stuck, ASH learns an Inverse Dynamics Model (IDM) from its own trajectories, and uses its IDM to extract supervision from relevant internet video. ASH uses unsupervised learning to identify key moments from large-scale internet video and retains them as long-term memory -- allowing it to tackle long-horizon problems. We evaluate ASH on two complementary environments demanding multi-hour planning: Pokemon Emerald, a turn-based RPG, and The Legend of Zelda: The Minish Cap, a real-time action-adventure game. In both games, behavioral cloning, retrieval-augmented and zero-shot foundation-model baselines plateau, while ASH sustains progression across our 8-hour evaluation. ASH reaches an average of $11.2/12$ milestones in Pokemon Emerald and $9.9/12$ in Legend of Zelda, while the strongest baseline gets stuck in both environments at an average of $6.5/12$ and $6.0/12$ milestones, respectively. We demonstrate that self-improving agents are a scalable recipe for long-horizon embodied learning.
Abstract:Long-horizon embodied tasks remain a fundamental challenge in AI, as current methods rely on hand-engineered rewards or action-labeled demonstrations, neither of which scales. We introduce ASH, an agentic system that learns an embodied policy from unlabeled, noisy internet video, without reward shaping or expert annotation. ASH follows a self-improvement loop; when it gets stuck, ASH learns an Inverse Dynamics Model (IDM) from its own trajectories, and uses its IDM to extract supervision from relevant internet video. ASH uses unsupervised learning to identify key moments from large-scale internet video and retains them as long-term memory -- allowing it to tackle long-horizon problems. We evaluate ASH on two complementary environments demanding multi-hour planning: Pokemon Emerald, a turn-based RPG, and The Legend of Zelda: The Minish Cap, a real-time action-adventure game. In both games, behavioral cloning, retrieval-augmented and zero-shot foundation-model baselines plateau, while ASH sustains progression across our 8-hour evaluation. ASH reaches an average of $11.2/12$ milestones in Pokemon Emerald and $9.9/12$ in Legend of Zelda, while the strongest baseline gets stuck in both environments at an average of $6.5/12$ and $6.0/12$ milestones, respectively. We demonstrate that self-improving agents are a scalable recipe for long-horizon embodied learning.
Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
arXiv:2605.14211 [cs.AI]
(or arXiv:2605.14211v3 [cs.AI] for this version)
Text integral sub licența deschisă a publicației · 208 cuvinte
Subiecte conexe
EmbodiedCLBench: Evaluating Continual Learning for Self-Evolving Embodied Agents
arXiv · 12 oct. 2026
[AINews] Reality Checks on AI News (Yegge shuts down Gas Town, Databricks’ +60% Astra cost)
Latent Space · 17 sept. 2026
2026 BAIR Graduate Showcase
BAIR Blog · 1 iul. 2026
Self-Evolving Embodied Agents via Skill-Harness Evolution
arXiv · 11 sept. 2026