新闻
Evidence-Type Competition: When Can Interventional Data Teach Language Models Causal Direction?
arXiv cs.AI · 发布于 · 阅读约3分钟
30秒读懂
- 发生了什么
- Research shows language models trained on interventional causal data still copy observational correlations at inference time, contradicting assumptions about causal learning.
- 为何重要
- Matters for engineers building systems requiring reliable causal reasoning, especially when training data mixes observational and interventional examples.
- 注意
- The findings use synthetic environments with Simpson's paradox; real-world applicability and scaling behavior beyond 0.93B parameters remain unclear.
收听本摘要
- language model
- reasoning
The Agent Architect
每周一个模式、一个权衡、一个生产事故案例。为构建智能体系统的人准备的每周简报。
每周一封邮件,一键退订。您的地址仅用于发送简报。