arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2306.02707cs.CLcs.LG
收藏

Orca:从 GPT-4 的复杂解释轨迹中进行渐进式学习

Orca: Progressive Learning from Complex Explanation Traces of GPT-4

Microsoft Research(微软研究院)

Subhabrata Mukherjee, Arindam Mitra, Ganesh Jawahar, Sahaj Agarwal, Hamid Palangi, Ahmed Awadallah

AI总结:

本文提出 130 亿参数模型 Orca,利用 GPT-4 的逐步解释轨迹和 ChatGPT 辅助进行渐进式模仿学习,在复杂零样本推理与专业考试中显著优于 Vicuna-13B 并接近 ChatGPT。

摘要:

近期研究聚焦于通过模仿学习增强较小模型的能力,其训练依据是大型基础模型(LFM)生成的输出。若干问题影响这些模型的质量:浅层 LFM 输出提供的模仿信号有限;小规模、同质化的训练数据;以及最突出的缺乏严格评估,导致高估小模型能力,因为它们往往只学会模仿风格,而没有学会 LFM 的推理过程。为应对这些挑战,我们开发了 Orca(我们正与法务团队合作,按照 LLaMA 发布政策公开发布模型权重差异文件,地址为 https://aka.ms/orca-lm),这是一个拥有 130 亿参数、学习模仿 LFM 推理过程的模型。Orca 从 GPT-4 的丰富信号中学习,包括解释轨迹、逐步思维过程以及其他复杂指令,并在 ChatGPT 的教师辅助下进行引导。为推动这种渐进式学习,我们利用大规模、多样化的模仿数据,并进行审慎采样和选择。在 Big-Bench Hard(BBH)等复杂零样本推理基准上,Orca 比 Vicuna-13B 等传统最先进指令微调模型高出 100% 以上,在 AGIEval 上高出 42%。此外,Orca 在 BBH 基准上达到与 ChatGPT 相当的水平;在 SAT、LSAT、GRE 和 GMAT 等专业与学术考试中,即便在没有 CoT 的零样本设置下,也表现出具有竞争力的性能(使用优化系统消息时差距为 4 分),但仍落后于 GPT-4。我们的研究表明,从逐步解释中学习——无论这些解释由人类还是更先进的 AI 模型生成——都是提升模型能力和技能的一条有前景的方向。

英文摘要:

Recent research has focused on enhancing the capability of smaller models through imitation learning, drawing on the outputs generated by large foundation models (LFMs). A number of issues impact the quality of these models, ranging from limited imitation signals from shallow LFM outputs; small scale homogeneous training data; and most notably a lack of rigorous evaluation resulting in overestimating the small model's capability as they tend to learn to imitate the style, but not the reasoning process of LFMs. To address these challenges, we develop Orca (We are working with our legal team to publicly release a diff of the model weights in accordance with LLaMA's release policy to be published at https://aka.ms/orca-lm), a 13-billion parameter model that learns to imitate the reasoning process of LFMs. Orca learns from rich signals from GPT-4 including explanation traces; step-by-step thought processes; and other complex instructions, guided by teacher assistance from ChatGPT. To promote this progressive learning, we tap into large-scale and diverse imitation data with judicious sampling and selection. Orca surpasses conventional state-of-the-art instruction-tuned models such as Vicuna-13B by more than 100% in complex zero-shot reasoning benchmarks like Big-Bench Hard (BBH) and 42% on AGIEval. Moreover, Orca reaches parity with ChatGPT on the BBH benchmark and shows competitive performance (4 pts gap with optimized system message) in professional and academic examinations like the SAT, LSAT, GRE, and GMAT, both in zero-shot settings without CoT; while trailing behind GPT-4. Our research indicates that learning from step-by-step explanations, whether these are generated by humans or more advanced AI models, is a promising direction to improve model capabilities and skills.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑