aiqbench analysis
This paper proposes CoDA, a loop in which a simulator generates robot training environments aimed at the policy's observed failures, while a vision-language critic diagnoses those failures and a reward-shaping method (Text2potential) supplies dense learning signals. The reported gains are large, such as out-of-distribution success rising from 5.3% to 27.6% on manipulation tasks, but they come from a few task families and benchmarks (including LIBERO-Pro), so generalization to other robots and real-world settings remains untested. Readers should watch whether targeted, failure-driven simulation can match the cost of simply collecting more data.
CoDA: Co-Evolution of Data Engines and Action Policies
arXiv:2610.22362v1 Announce Type: cross Abstract: Generative simulation can automate robot environment construction, but physically valid environments are not necessarily useful for training. The challenge is to identify environments that address a policy's failures and supervision that helps it overcome them. We introduce CoDA, a framework in which a simulation data engine and an action policy co-evolve through a shared heterogeneous memory of environment construction, policy failures, and learning progress. The memory compiles validated diagnoses from a vision-language critic into generation programs that recreate difficulties and training programs that help the policy overcome them. We further propose Text2potential to convert these diagnoses into staged potentials for action-chunk value learning, providing dense supervision for policy updates while preserving the task's optimal policy under standard shaping assumptions. Measured learning progress guides further generation and program refinement. Across four manipulation task families, CoDA improves out-of-distribution success from 5.3% after supervised fine-tuning to 27.6%, exceeding the strongest closed-loop baseline by 9.5 percentage points with the same generation budget. On LIBERO-Pro, 15 rounds of generation and training raise success from 25% to 84% under position perturbations and from 1% to 76% under instruction perturbations.
Abstract:Generative simulation can automate robot environment construction, but physically valid environments are not necessarily useful for training. The challenge is to identify environments that address a policy's failures and supervision that helps it overcome them. We introduce CoDA, a framework in which a simulation data engine and an action policy co-evolve through a shared heterogeneous memory of environment construction, policy failures, and learning progress. The memory compiles validated diagnoses from a vision-language critic into generation programs that recreate difficulties and training programs that help the policy overcome them. We further propose Text2potential to convert these diagnoses into staged potentials for action-chunk value learning, providing dense supervision for policy updates while preserving the task's optimal policy under standard shaping assumptions. Measured learning progress guides further generation and program refinement. Across four manipulation task families, CoDA improves out-of-distribution success from 5.3% after supervised fine-tuning to 27.6%, exceeding the strongest closed-loop baseline by 9.5 percentage points with the same generation budget. On LIBERO-Pro, 15 rounds of generation and training raise success from 25% to 84% under position perturbations and from 1% to 76% under instruction perturbations.
Robotics (cs.RO); Artificial Intelligence (cs.AI)
arXiv:2610.22362 [cs.RO]
(or arXiv:2610.22362v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2610.22362 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 279 words
Related coverage
Article: An Evolutionary Architecture Pattern for Managing AI’s Pace of Change
InfoQ · Jul 27, 2026
Enabling Evolutionary Database Development: Database branching with Lakebase, the conclusion
Databricks · Jun 12, 2026
California Government Technology Summit 2026
events.govtech.com · Aug 7, 2026
From weeks to minutes: How Formula 1® uses agentic AI on AWS to accelerate data operations
AWS · Aug 3, 2026