• Search Research Projects
  • Search Researchers
  • How to Use
  1. Back to project page

2023 Fiscal Year Research-status Report

Is "2019" Earlier than "2022"? Unsupervised Temporal Representation Learning via Contrastive Learning.

Research Project

Project/Area Number 23K16946
Research InstitutionKyoto University

Principal Investigator

程 飛  京都大学, 情報学研究科, 特定助教 (70801570)

Project Period (FY) 2023-04-01 – 2025-03-31
KeywordsTemporal reasnoning / Large language model / Common sense reasoning / Numerical reasoning / Contrastive learning
Outline of Annual Research Achievements

My research aims to improve the temporal reasoning capabilitie (common sense + numerical reasoning) of large language models (LLMs). I mainly achieve two things in this year: (1) Inspired by Stanford University's Alpaca model , I proposed a method to distill Japanese knowledge from powerful GPT-4 to improve open LLMs’ capability of Japanese common sense reasoning. (2) Developing an QA benchmark that can assess LLMs’ reasoning capabilities across eight dimensions: common sense, mathematical operations, writing, etc. Our model leverages state-of-the-art GPT-4 as a judge to assess LLMs’ outputs.

Current Status of Research Progress
Current Status of Research Progress

2: Research has progressed on the whole more than it was originally planned.

Reason

According to our plan, we spent the first year preparing the design of the model and conducting some preliminary experiments. Our idea is to leverage contrastive learning to train LLMs on corpora containing rich temporal and numerical expressions, to achieve better quality of temporal representations. The actual progress was quite smooth; based on the latest LLM research worldwide this year, we distill a small amount of high-quality instruction data (related to common sense, numerical reasoning) from powerful GPT-4. We then train open LLMs including OpenCalm, LLaMA1, and LLaMA2 on this data. All three models achieved improvements in reasoning performance, including common sense reasoning and numerical reasoning. The paper is accepted by the international conference LREC-COLING 2024.

Strategy for Future Research Activity

In the latest experiment, we find that a small amount of data distilled from GPT-4 could significantly improve LLMs' capability of common sense and numerical reasoning. This leads us to consider whether we could drive GPT-4 to intentionally create pairs of high-quality and low-quality temporal textual data directly. These pairs could serve as positive and negative examples for our contrastive learning to optimize the representations of the latest open Japanese LLMs such as LLM-jp 13B, Swallow, etc. This approach would allow us to avoid handling large amounts of raw text and extracting contrastive learning targets with low relevance from raw corpora. Our goal is to directly optimize the temporal representations and also incorporate temporal reasoning into our QA benchmark.

  • Research Products

    (4 results)

All 2024 2023 Other

All Int'l Joint Research (1 results) Journal Article (2 results) (of which Int'l Joint Research: 1 results,  Peer Reviewed: 2 results,  Open Access: 1 results) Remarks (1 results)

  • [Int'l Joint Research] Peking University/Xiaomi AI Lab(中国)

    • Country Name
      CHINA
    • Counterpart Institution
      Peking University/Xiaomi AI Lab
  • [Journal Article] Rapidly Developing High-quality Instruction Data and Evaluation Benchmark for Large Language Models with Minimal Human Effort: A Case Study on Japanese2024

    • Author(s)
      Yikun Sun, Zhen Wan, Nobuhiro Ueda, Sakiko Yahata, Fei Cheng, Chenhui Chu, Sadao Kurohashi
    • Journal Title

      Proceedings of The 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (COLING-LREC 2024)

      Volume: v1 Pages: 0,0

    • Peer Reviewed
  • [Journal Article] ComSearch: Equation Searching with Combinatorial Strategy for Solving Math Word Problems with Weak Supervision2023

    • Author(s)
      Qianying Liu, Wenyu Guan, Jianhao Shen, Fei Cheng, Sadao Kurohashi
    • Journal Title

      Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics (EACL 2023)

      Volume: v1 Pages: 2549, 2562

    • DOI

      10.18653/v1/2023.eacl-main.186

    • Peer Reviewed / Open Access / Int'l Joint Research
  • [Remarks] QA Benchmark for evaluating Japanese LLMs

    • URL

      https://github.com/ku-nlp/ja-vicuna-qa-benchmark

URL: 

Published: 2024-12-25  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi