Piccolo: Co-Evolving Decision Models and Data through Recursive Self-Improvement
We present Piccolo, a family of multimodal decision models (0.8B to 27B) that coevolves model capabilities and training data via Agentic Data Engineering and recursive self-improvement (RSI). The models predict probabilities over candidate answers for general-purpose text and visual judgments, agent harnesses, data workflows, and interactive games. Piccolo-27B tops the Decision Index Vision Full leaderboard with a score of 75.59, while Piccolo-9B and Piccolo-4B achieve 68.35 and 67.46, respectively. All three rank first on the Private benchmark in their respective model-size categories. Research agents analyze model behavior, construct training data, conduct experiments, and use the results to refine subsequent proposals. Semantic slices and a shared research memory connect these actions across the model family's development. With training recipes comprising hundreds of thousands of examples, data selection and feedback-driven refinement are central to the self-improvement process. Our large-scale RSI exploration yields three counter-intuitive insights for decision-model development: (1) Backbone-recipe co-adaptation outweighs raw model recency: earlier backbones can surpass newer generations on visual tasks under identical recipes; (2) Targeted data curation enables 4B-9B models to rival 27B baselines on specialized workflows; and (3) Continuous loss reduction can decouple from and even degrade discrete decision accuracy. Together with the released weights and tools from the Data-Juicer and AgentScope ecosystems, Piccolo provides an actionable blueprint for building frontier decision models and scaling data-driven RSI.
Authors
- Bolin Ding (ORCID: https://orcid.org/0000-0003-1535-9692)
- Weirui Kuang (ORCID: https://orcid.org/0000-0001-8544-0221)
- Mengsu Ding
- Xiaoming Qin (ORCID: https://orcid.org/0000-0001-8408-3998)
- Xinji Mai (ORCID: https://orcid.org/0009-0003-4596-5391)
- Qirui Jiao
- Zhenqing Ling (ORCID: https://orcid.org/0009-0007-7834-8444)
- Jingren Zhou
- Daoyuan Chen
- Zhuo Zhang
- Yaliang Li
- Zhaoyang Liu
- Xuchen Pan
- Yuexiang Xie
- Cheng Chen
- Jiaji Deng
- Haowen Gao
Publication Details
- Journal
- arXiv (Cornell University)
- Published
- 2026-11-18
- DOI
- https://doi.org/10.48550/arxiv.2611.24651
- Primary Topic
- Multimodal Machine Learning Applications
- Type
- preprint