CROCODIL：通过奖励机制抑制大模型过度编辑代码
This is a weird behavior in coding models and something worth looking into.
揭示了多模型协作编码中的关键缺陷，CROCODIL 的乘法奖励机制提供了可复现的优化思路，值得 Agent 开发者关注。
This is a weird behavior in coding models and something worth looking into.
这是代码模型中的一种奇怪行为，值得深入研究。
It turns that some models over-edit code that another models wrote.
事实证明，某些模型会过度编辑其他模型编写的代码。
There is a high chance that your repo now has commits from more than one model, and that changes how each of them edits.
你的仓库现在很可能包含来自多个模型的提交，这会改变每个模型的编辑方式。
Researchers measured what happens when one model edits code another model wrote. Different training data produces different stylistic preferences, and models make more edits, often excessive ones, on foreign code than on their own.
研究人员测量了一个模型编辑另一个模型所写代码时会发生什么。不同的训练数据会产生不同的风格偏好，模型对外部代码的编辑往往比对自己代码的更多，且常常是过度的。
CROCODIL is a post-training framework that reduces that behavior. A similarity reward penalizes large changes and an execution reward scores build and test success, and the two are multiplied rather than added. That product stops the policy from shrinking edits by simply failing the task.
CROCODIL 是一种后训练框架，用于减少这种行为。相似度奖励会对大幅更改进行惩罚，执行奖励则评估构建和测试的成功率，两者相乘而非相加。这种乘积机制防止策略通过简单地使任务失败来缩减编辑量。
Paper: https://academy.dair.ai/papers/crocodil-cross-model-code-editing-with-llms-2609.03894
论文：https://academy.dair.ai/papers/crocodil-cross-model-code-editing-with-llms-2609.03894
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力