Flacuna：利用FLAN Fine-Tuning釋放維庫納的問題解決能力

Flacuna: Unleashing the Problem Solving Power of Vicuna using FLAN Fine-Tuning

July 5, 2023

作者: Deepanway Ghosal, Yew Ken Chia, Navonil Majumder, Soujanya Poria

cs.AI

摘要

最近，INSTRUCTEVAL 的發布為利用編碼器-解碼器或僅解碼器架構的大型語言模型（LLMs）的表現提供了寶貴的見解。有趣的是，儘管四年前推出，基於 T5 的 LLMs，如 FLAN-T5，仍然在需要一般問題解決能力的任務上優於最新的基於解碼器的 LLMs，如 LLAMA 和 VICUNA。這種性能差異可以歸因於三個關鍵因素：（1）預訓練數據，（2）骨幹架構，以及（3）指令數據集。在這份技術報告中，我們的主要重點是通過利用基於 LLAMA 的大型語言模型 VICUNA，該模型已在 ChatGPT 對話上進行了微調，來研究第三個因素的影響。為了實現這一目標，我們使用了一個名為 FLANMINI 的定製指令數據集對 VICUNA 進行了微調。該數據集包括 FLAN 這個大規模指令數據集的子集，以及從 ChatGPT/GPT-4 衍生的各種與代碼相關的數據集和對話數據集。這個數據集包含大量需要解決問題技能的任務。我們的實驗結果強烈表明，我們的模型 FLACUNA 的增強問題解決能力是通過在 FLAN 數據集上對 VICUNA 進行微調實現的，從而在 INSTRUCTEVAL 的眾多基準數據集上實現了顯著改進。FLACUNA 可在 https://huggingface.co/declare-lab/flacuna-13b-v1.0 公開獲取。

English

Recently, the release of INSTRUCTEVAL has provided valuable insights into the performance of large language models (LLMs) that utilize encoder-decoder or decoder-only architecture. Interestingly, despite being introduced four years ago, T5-based LLMs, such as FLAN-T5, continue to outperform the latest decoder-based LLMs, such as LLAMA and VICUNA, on tasks that require general problem-solving skills. This performance discrepancy can be attributed to three key factors: (1) Pre-training data, (2) Backbone architecture, and (3) Instruction dataset. In this technical report, our main focus is on investigating the impact of the third factor by leveraging VICUNA, a large language model based on LLAMA, which has undergone fine-tuning on ChatGPT conversations. To achieve this objective, we fine-tuned VICUNA using a customized instruction dataset collection called FLANMINI. This collection includes a subset of the large-scale instruction dataset known as FLAN, as well as various code-related datasets and conversational datasets derived from ChatGPT/GPT-4. This dataset comprises a large number of tasks that demand problem-solving skills. Our experimental findings strongly indicate that the enhanced problem-solving abilities of our model, FLACUNA, are obtained through fine-tuning VICUNA on the FLAN dataset, leading to significant improvements across numerous benchmark datasets in INSTRUCTEVAL. FLACUNA is publicly available at https://huggingface.co/declare-lab/flacuna-13b-v1.0.