符号调整提高了语言模型中的上下文学习。

摘要

我们提出了符号微调 - 在上下文输入-标签对上微调语言模型，其中自然语言标签（例如，“正面/负面情感”）被任意符号（例如，“foo/bar”）替换。符号微调利用了这样的直觉，即当模型无法使用说明或自然语言标签来解决任务时，必须通过学习输入-标签映射来实现。我们在Flan-PaLM模型上进行了符号微调的实验，参数量高达540B，并观察到在各种设置下的好处。首先，符号微调提升了在未见过的上下文学习任务上的性能，并且对于指令不明确或没有自然语言标签的提示更加稳健。其次，经过符号微调的模型在算法推理任务上表现更加强大，List Functions基准测试上性能提高了高达18.2%，Simple Turing Concepts基准测试上性能提高了高达15.3%。最后，经过符号微调的模型在跟随上下文中呈现的反转标签方面显示出了很大的改进，这意味着它们更能够利用上下文信息来覆盖先前的语义知识。

English

We present symbol tuning - finetuning language models on in-context input-label pairs where natural language labels (e.g., "positive/negative sentiment") are replaced with arbitrary symbols (e.g., "foo/bar"). Symbol tuning leverages the intuition that when a model cannot use instructions or natural language labels to figure out a task, it must instead do so by learning the input-label mappings. We experiment with symbol tuning across Flan-PaLM models up to 540B parameters and observe benefits across various settings. First, symbol tuning boosts performance on unseen in-context learning tasks and is much more robust to underspecified prompts, such as those without instructions or without natural language labels. Second, symbol-tuned models are much stronger at algorithmic reasoning tasks, with up to 18.2% better performance on the List Functions benchmark and up to 15.3% better performance on the Simple Turing Concepts benchmark. Finally, symbol-tuned models show large improvements in following flipped-labels presented in-context, meaning that they are more capable of using in-context information to override prior semantic knowledge.

符号调整提高了语言模型中的上下文学习。

Symbol tuning improves in-context learning in language models

摘要

Support