Joey-1 在讽刺检测任务上表现更优
摘要
有趣的是，即便是 Joey-1（一个 2B 模型）在讽刺检测任务上也比我们的模型表现更好。 需要注意的是，Baseten 的三个模型都是在更大的数据集上训练的——621 万个样本，约为我们的 53 倍。
摘要由模型根据原文撰写，以原文为准。
正文 · 原文
Interestingly, even Joey-1 (a 2B model) performed better than our model at sarcasm detection tasks.
Note that all three of Baseten's models were trained over a much larger dataset - 6.21 million examples, which is ~53x ours.
来源：alphaXiv · x.com
引用原文
askalphaxiv. Interestingly, even Joey-1 (a 2B model) performed better than our model at sarcasm detection tasks.[EB/OL]. (2026-10-31)[访问日期]. https://x.com/askalphaxiv/status/2116199952565318143.