从自产数据中泛化：超越人类约束的模型训练框架

Generalising from Self-Produced Data: Model Training Beyond Human Constraints

AI 导读

该论文提出一种让 AI 模型通过与环境的直接交互自主生成并验证新知识的框架，核心是一个不可博弈的无界数值奖励（如磁盘占用空间或粉丝数），无需人类基准即可引导学习。AI 智能体迭代生成策略和可执行代码以最大化该指标，成功结果用于自我再训练和增量泛化，并通过 GRPO 支持微调以缓解模型崩溃和冷启动问题。系统采用模块化智能体完成环境分析、策略生成与代码合成。

正文 · 原文

View PDF HTML (experimental)

Abstract:Current large language models (LLMs) are constrained by human-derived training data and limited by a single level of abstraction that impedes definitive truth judgments. This paper introduces a novel framework in which AI models autonomously generate and validate new knowledge through direct interaction with their environment. Central to this approach is an unbounded, ungamable numeric reward - such as annexed disk space or follower count - that guides learning without requiring human benchmarks. AI agents iteratively generate strategies and executable code to maximize this metric, with successful outcomes forming the basis for self-retraining and incremental generalisation. To mitigate model collapse and the warm start problem, the framework emphasizes empirical validation over textual similarity and supports fine-tuning via GRPO. The system architecture employs modular agents for environment analysis, strategy generation, and code synthesis, enabling scalable experimentation. This work outlines a pathway toward self-improving AI systems capable of advancing beyond human-imposed constraints toward autonomous general intelligence.

Comments:	16 pages, 2 figures
Subjects:	Artificial Intelligence (cs.AI)
Cite as:	arXiv:2504.04711 [cs.AI]
	(or arXiv:2504.04711v3 [cs.AI] for this version)
	https://doi.org/10.48550/arXiv.2504.04711 arXiv-issued DOI via DataCite

Submission history

From: Akira Rafhael Janson Pattirane [view email]
[v1] Mon, 7 Apr 2025 03:48:02 UTC (481 KB)
[v2] Wed, 30 Sep 2026 12:09:11 UTC (323 KB)
[v3] Thu, 15 Oct 2026 06:47:24 UTC (323 KB)

来源：arXiv cs.AI · arxiv.org