Presentation 2023-09-06
LoRA Tuning Conversational Japanese Large Language Models using Japanese Instruction Dataset
Masahiro Suzuki, Masanori Hirano, Hiroki Sakaji,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this study, we performed LoRA tuning on large language models (LLM) based on both Japanese and English using Japanese instruction tuning and evaluated these models from both quantitative and qualitative perspectives. As a result of the evaluation, the effectiveness of tuning with Japanese instruction data was confirmed. Furthermore, we clarified the challenges in large-scale language models and language resources in Japanese, such as the need for evaluation using a wide range of instruction data and the actual output strings of the models.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Large Language Model (LLM) / Japanese / Instruction Tuning
Paper # NLC2023-1
Date of Issue 2023-08-30 (NLC)

Conference Information
Committee NLC
Conference Date 2023/9/6(2days)
Place (in Japanese) (See Japanese page)
Place (in English) Osaka Metropolitan University. Nakamozu Campus.
Topics (in Japanese) (See Japanese page)
Topics (in English) The 20th Text Analytics Symposium
Chair Mitsuo Yoshida(Univ. of Tsukuba)
Vice Chair Hiroki Sakaji(Univ. of Tokyo) / Takeshi Kobayakawa(NHK)
Secretary Hiroki Sakaji(rinna) / Takeshi Kobayakawa(Hiroshima Univ. of Economics)
Assistant Kanjin Takahashi(Sansan) / Yasuhiro Ogawa(Nagoya Univ.)

Paper Information
Registration To Technical Committee on Natural Language Understanding and Models of Communication
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) LoRA Tuning Conversational Japanese Large Language Models using Japanese Instruction Dataset
Sub Title (in English)
Keyword(1) Large Language Model (LLM)
Keyword(2) Japanese
Keyword(3) Instruction Tuning
1st Author's Name Masahiro Suzuki
1st Author's Affiliation The University of Tokyo(UTokyo)
2nd Author's Name Masanori Hirano
2nd Author's Affiliation The University of Tokyo(UTokyo)
3rd Author's Name Hiroki Sakaji
3rd Author's Affiliation The University of Tokyo(UTokyo)
Date 2023-09-06
Paper # NLC2023-1
Volume (vol) vol.123
Number (no) NLC-176
Page pp.pp.1-6(NLC),
#Pages 6
Date of Issue 2023-08-30 (NLC)