OpenResearch 训练 4B 决策模型
We trained a 4B decision model on over 117k decision-making examples with OpenResearch
摘要
我们用 OpenResearch 在超过 117k 个决策示例上训练了一个 4B 决策模型 我们的模型在除讽刺检测外的所有领域表现几乎与 Jev 完全相同 在评估中，我们的模型能够泛化到其他未见过的任务，如仇恨言论严重程度、情绪检测和意图分类，但讽刺仍是一个常见的失败模式
摘要由模型根据原文撰写，以原文为准。
正文 · 原文
We trained a 4B decision model on over 117k decision-making examples with OpenResearch
Our model performs practically identically to Jev in all areas except one: sarcasm detection
During evals, our model was able to generalize to other, unseen tasks like hate-speech severity, emotion detection, and intent classification, but sarcasm remained a common failure mode
来源：alphaXiv · x.com
引用原文
askalphaxiv. We trained a 4B decision model on over 117k decision-making examples with OpenResearch[EB/OL]. (2026-10-31)[访问日期]. https://x.com/askalphaxiv/status/2116199951835582507.