Skip to content

Instantly share code, notes, and snippets.

@mlabonne
Created February 13, 2024 17:53
Show Gist options
  • Save mlabonne/f1fb8cf80819338fdad49ad395626040 to your computer and use it in GitHub Desktop.
Save mlabonne/f1fb8cf80819338fdad49ad395626040 to your computer and use it in GitHub Desktop.
Model AGIEval GPT4All TruthfulQA Bigbench Average
Monarch-7B-slerp 45.13 77.09 78.63 49.56 62.6

AGIEval

Task Version Metric Value Stderr
agieval_aqua_rat 0 acc 27.17 ± 2.80
acc_norm 25.20 ± 2.73
agieval_logiqa_en 0 acc 39.94 ± 1.92
acc_norm 40.40 ± 1.92
agieval_lsat_ar 0 acc 25.22 ± 2.87
acc_norm 24.35 ± 2.84
agieval_lsat_lr 0 acc 53.53 ± 2.21
acc_norm 54.90 ± 2.21
agieval_lsat_rc 0 acc 65.43 ± 2.91
acc_norm 65.43 ± 2.91
agieval_sat_en 0 acc 77.67 ± 2.91
acc_norm 77.67 ± 2.91
agieval_sat_en_without_passage 0 acc 42.72 ± 3.45
acc_norm 41.75 ± 3.44
agieval_sat_math 0 acc 33.18 ± 3.18
acc_norm 31.36 ± 3.14

Average: 45.13%

GPT4All

Task Version Metric Value Stderr
arc_challenge 0 acc 66.81 ± 1.38
acc_norm 68.43 ± 1.36
arc_easy 0 acc 86.28 ± 0.71
acc_norm 80.01 ± 0.82
boolq 1 acc 87.13 ± 0.59
hellaswag 0 acc 70.33 ± 0.46
acc_norm 87.57 ± 0.33
openbookqa 0 acc 39.60 ± 2.19
acc_norm 49.80 ± 2.24
piqa 0 acc 82.92 ± 0.88
acc_norm 84.93 ± 0.83
winogrande 0 acc 81.77 ± 1.09

Average: 77.09%

TruthfulQA

Task Version Metric Value Stderr
truthfulqa_mc 1 mc1 63.65 ± 1.68
mc2 78.63 ± 1.37

Average: 78.63%

Bigbench

Task Version Metric Value Stderr
bigbench_causal_judgement 0 multiple_choice_grade 57.89 ± 3.59
bigbench_date_understanding 0 multiple_choice_grade 61.52 ± 2.54
bigbench_disambiguation_qa 0 multiple_choice_grade 55.43 ± 3.10
bigbench_geometric_shapes 0 multiple_choice_grade 22.84 ± 2.22
exact_str_match 0.84 ± 0.48
bigbench_logical_deduction_five_objects 0 multiple_choice_grade 33.80 ± 2.12
bigbench_logical_deduction_seven_objects 0 multiple_choice_grade 23.71 ± 1.61
bigbench_logical_deduction_three_objects 0 multiple_choice_grade 58.33 ± 2.85
bigbench_movie_recommendation 0 multiple_choice_grade 53.80 ± 2.23
bigbench_navigate 0 multiple_choice_grade 56.50 ± 1.57
bigbench_reasoning_about_colored_objects 0 multiple_choice_grade 68.10 ± 1.04
bigbench_ruin_names 0 multiple_choice_grade 56.47 ± 2.35
bigbench_salient_translation_error_detection 0 multiple_choice_grade 40.58 ± 1.56
bigbench_snarks 0 multiple_choice_grade 74.03 ± 3.27
bigbench_sports_understanding 0 multiple_choice_grade 74.95 ± 1.38
bigbench_temporal_sequences 0 multiple_choice_grade 53.40 ± 1.58
bigbench_tracking_shuffled_objects_five_objects 0 multiple_choice_grade 22.88 ± 1.19
bigbench_tracking_shuffled_objects_seven_objects 0 multiple_choice_grade 19.43 ± 0.95
bigbench_tracking_shuffled_objects_three_objects 0 multiple_choice_grade 58.33 ± 2.85

Average: 49.56%

Average score: 62.6%

Elapsed time: 02:28:34

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment