Zobrazeno 1 - 3
of 3
pro vyhledávání: '"Sun, Zhouhao"'
Although achieving promising performance, recent analyses show that current generative large language models (LLMs) may still capture dataset biases and utilize them for generation, leading to poor generalizability and harmfulness of LLMs. However, d
Externí odkaz:
http://arxiv.org/abs/2408.12942
Large language models (LLMs) have achieved significant performance in various natural language reasoning tasks. However, they still struggle with performing first-order logic reasoning over formal logical theories expressed in natural language. This
Externí odkaz:
http://arxiv.org/abs/2404.01677
Through pretraining on a corpus with various sources, Large Language Models (LLMs) have gained impressive performance. However, the impact of each component of the pretraining corpus remains opaque. As a result, the organization of the pretraining co
Externí odkaz:
http://arxiv.org/abs/2402.11537