aaaaaaapluto/

1b-model-eval

A comprehensive G-Eval (Liu et al., EMNLP 2023) benchmark of 16 small language models under 2B parameters, covering 6 dimensions across 33 test items. Judge model: DeepSeek-v4-pro

	Files




	

File Description

nightly_raw_.json Raw outputs for each model — 33 generated items across 6 dimensions

nightly_eval_results.json Judge scores and detailed reasoning for all 16 models

summary.csv One-row-per-model score… See the full description on the dataset page: https://huggingface.co/datasets/aaaaaaapluto/1b-model-eval.