SOLAR-10.7B-Instruct-v1.0-laser

This version of Solar-10.7B was lasered and perplexity was calculated against gsm8k.

• Initial Model Perplexity: 12.865185737609863
• New baseline perplexity: 12.554274559020996

The laser process decreased perplexity by 2.41%

Model	AGIEval	GPT4All	TruthfulQA	Bigbench	Average
SOLAR-10.7B-Instruct-v1.0-laser	46.9	74.99	70.64	43.74	59.07

AGIEval

Task	Version	Metric	Value		Stderr
agieval_aqua_rat	0	acc	29.53	±	2.87
		acc_norm	28.35	±	2.83
agieval_logiqa_en	0	acc	39.78	±	1.92
		acc_norm	40.55	±	1.93
agieval_lsat_ar	0	acc	23.04	±	2.78
		acc_norm	21.30	±	2.71
agieval_lsat_lr	0	acc	51.18	±	2.22
		acc_norm	51.76	±	2.21
agieval_lsat_rc	0	acc	66.54	±	2.88
		acc_norm	66.91	±	2.87
agieval_sat_en	0	acc	78.16	±	2.89
		acc_norm	78.16	±	2.89
agieval_sat_en_without_passage	0	acc	50.97	±	3.49
		acc_norm	50.00	±	3.49
agieval_sat_math	0	acc	42.73	±	3.34
		acc_norm	38.18	±	3.28

Average: 46.9%

GPT4All

Task	Version	Metric	Value		Stderr
arc_challenge	0	acc	60.84	±	1.43
		acc_norm	63.99	±	1.40
arc_easy	0	acc	83.59	±	0.76
		acc_norm	81.44	±	0.80
boolq	1	acc	87.58	±	0.58
hellaswag	0	acc	68.11	±	0.47
		acc_norm	85.77	±	0.35
openbookqa	0	acc	35.40	±	2.14
		acc_norm	48.40	±	2.24
piqa	0	acc	80.58	±	0.92
		acc_norm	80.74	±	0.92
winogrande	0	acc	77.03	±	1.18

Average: 74.99%

TruthfulQA

Task	Version	Metric	Value		Stderr
truthfulqa_mc	1	mc1	55.45	±	1.74
		mc2	70.64	±	1.49

Average: 70.64%

Bigbench

Task	Version	Metric	Value		Stderr
bigbench_causal_judgement	0	multiple_choice_grade	57.37	±	3.60
bigbench_date_understanding	0	multiple_choice_grade	62.87	±	2.52
bigbench_disambiguation_qa	0	multiple_choice_grade	35.66	±	2.99
bigbench_geometric_shapes	0	multiple_choice_grade	33.15	±	2.49
		exact_str_match	0.00	±	0.00
bigbench_logical_deduction_five_objects	0	multiple_choice_grade	26.20	±	1.97
bigbench_logical_deduction_seven_objects	0	multiple_choice_grade	19.71	±	1.50
bigbench_logical_deduction_three_objects	0	multiple_choice_grade	45.00	±	2.88
bigbench_movie_recommendation	0	multiple_choice_grade	39.00	±	2.18
bigbench_navigate	0	multiple_choice_grade	51.20	±	1.58
bigbench_reasoning_about_colored_objects	0	multiple_choice_grade	53.90	±	1.11
bigbench_ruin_names	0	multiple_choice_grade	40.18	±	2.32
bigbench_salient_translation_error_detection	0	multiple_choice_grade	39.98	±	1.55
bigbench_snarks	0	multiple_choice_grade	63.54	±	3.59
bigbench_sports_understanding	0	multiple_choice_grade	68.36	±	1.48
bigbench_temporal_sequences	0	multiple_choice_grade	65.20	±	1.51
bigbench_tracking_shuffled_objects_five_objects	0	multiple_choice_grade	22.48	±	1.18
bigbench_tracking_shuffled_objects_seven_objects	0	multiple_choice_grade	18.46	±	0.93
bigbench_tracking_shuffled_objects_three_objects	0	multiple_choice_grade	45.00	±	2.88

Average: 43.74%

Average score: 59.07%

Elapsed time: 02:33:24