The 7th LLM-jp Meeting

The 7th LLM-jp meeting was held on January 22, 2024 from 14:00 to 17:30 at National Institute of Informatics and available online.


  • Self-improvement loops from observational data, Grounding LLMs with causal inference methods (Sannai) [Material 1] [Material 2]
  • Comparative analysis of LLM human evaluation and GPT evaluation using ichikara-instruction (Sekine) [Material]
  • Kotoba Technologies voice-based model and Mamba architecture (Kojima and Kasai) [Material]
  • Large language model, Swallow (TokyoTech LLM members) [Material]
  • Safety Working Group (Sekine) [Material]
  • Corpus Construction Working Group (Kawahara) [Material]
  • Evaluation and Tuning Working Group (Miyao) [Material]
  • Model Building Working Group (Suzuki) [Material]


35 on-site and about 145 online