Marcus2112's picture
Create README.md
21ecff9 verified
metadata
datasets:
  - Marcus2112/minipile_density-proportioned
language:
  - en
base_model:
  - EleutherAI/pythia-160m-deduped
Benchmark Measure 160M Density 160M Density 2 Epochs Percentage Difference in Means
ARC-Challenge acc 0.1920 ± 0.0115 0.1894 ± 0.0115 -1.3542
MMLU acc 0.2295 ± 0.0035 0.2295 ± 0.0035 0.0000
HellaSwag acc 0.2604 ± 0.0044 0.2568 ± 0.0044 -1.3825
WinoGrande acc 0.5201 ± 0.0140 0.5012 ± 0.0141 -3.6339
Lambada (OpenAI) acc 0.0000 ± 0.0000 0.0000 ± 0.0000 -
Lambada (OpenAI) perplexity 2099002.0912 ± 170652.6222 1587737.3755 ± 121555.3148 -24.3575
Lambada (Std) acc 0.0000 ± 0.0000 0.0000 ± 0.0000 -
Lambada (Std) perplexity 13347273.6076 ± 1997894.6360 8366924.7603 ± 713077.3579 -37.3136
BLiMP acc 0.5501 ± 0.0017 0.5378 ± 0.0017 -2.2360