我所认知或定义的natural learning是一种依托任何文本或多模态上下文进行自然学习（广义权重更新）的机器学习范式。

1、从自然文本反馈中学习

<aside> 💡

Training Language Models with Language Feedback, Apr. 2022.

RL4F: Generating Natural Language Feedback with Reinforcement Learning for Repairing Model Outputs, ACL 2023.

On Improving Summarization Factual Consistency from Natural Language Feedback, ACL 2023.

Learning from Natural Language Feedback, TMLR 2024.

• Improving Code Generation by Training with Natural Language Feedback, Mar. 2023.
• Training Language Models with Language Feedback at Scale, Mar. 2023.

Test-Time Preference Optimization: On-the-Fly Alignment via Iterative Textual Feedback, Jan. 2025.

Text2Grad: Reinforcement learning from natural language feedback, May 2025.

Language models can learn from verbal feedback without scalar rewards, Sep. 2025.

Efficiently Aligning Language Models with Online Natural Language Feedback, May 2026.

Learning with Rare Success but Rich Feedback via Reflection-Enhanced Self-Distillation, May 2026.

</aside>