README.md

metadata

language:
  - en
  - ko
tags:
  - generation
license: apache-2.0
model-index:
  - name: task_1
    results:
      - task:
          type: natural-language-generation
        dataset:
          type: hellaswag
          name: hellaswag(10 shots)
        metrics:
          - type: acc_norm
            value: 27.7
  - name: task_2
    results:
      - task:
          type: natural-language-generation
        dataset:
          type: ARC
          name: ARC(25 shots)
        metrics:
          - type: acc_norm
            value: 23.8
  - name: task_3
    results:
      - task:
          type: natural-language-generation
        dataset:
          type: MMLU
          name: MMLU(5 shots)
        metrics:
          - type: acc
            value: 24.9
  - name: task_4
    results:
      - task:
          type: natural-language-generation
        dataset:
          type: TruthfulQA
          name: TruthfulQA(0 shots)
        metrics:
          - type: mc2
            value: 46.5

Pretrained GPT2 with expanded n_ctx up to 2048(also with expanded embedding dimension to 1536) in Korean.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	24.27
ARC (25-shot)	21.16
HellaSwag (10-shot)	28.11
MMLU (5-shot)	26.56
TruthfulQA (0-shot)	42.06
Winogrande (5-shot)	49.09
GSM8K (5-shot)	0.0
DROP (3-shot)	2.89