Dolci-Think-SFT-32B-500k-DE

499,430 samples from lightonai/Dolci-Think-SFT-32B-500k-DE

Category Distribution

Category	Count	%	Mean Tokens	Median	Std	Min	Max	Total Tokens
math	184,389	36.92%	24,380.1	23,947	7,455.6	1,384	45,963	4,495,421,714
code	169,846	34.01%	9,341.0	5,813	8,828.3	345	45,339	1,586,538,257
instruction-following	80,964	16.21%	3,085.0	1,748	3,844.4	41	43,329	249,770,081
science	25,462	5.10%	11,947.5	10,623	7,339.0	1,031	41,314	304,207,310
safety	19,490	3.90%	1,211.9	1,105	569.9	147	16,827	23,620,516
general-chat	18,134	3.63%	4,286.4	2,951	3,991.5	47	52,303	77,729,474
structured-data	1,145	0.23%	2,278.2	1,706	1,957.9	247	15,158	2,608,505

Token Count Statistics

Stat	Value
Count	499,430
Mean	13,495.2
Std	11,372.2
Min	41
25%	2,506
Median	10,893
75%	22,279
90%	30,764
95%	33,874
99%	37,798
Max	52,303
Total	6,739,895,857

Dataset to Category Mapping

Category	Original Dataset	Count	%	Mean Tokens
code	Dolci-Think-Python	121,928	24.41%	5,743.4
code	Nemotron Post-Training	27,328	5.47%	15,364.9
code	OpenThoughts3 (code)	20,590	4.12%	22,650.2
general-chat	WildChat (Tulu v3.9)	9,738	1.95%	4,393.8
general-chat	WildChat	7,597	1.52%	4,368.5
general-chat	OpenAssistant (OASST1)	799	0.16%	2,197.0
instruction-following	Dolci-Precise-IF	53,708	10.75%	2,104.6
instruction-following	SYNTHETIC-2 (IF/QwQ)	27,256	5.46%	5,016.7
math	OpenThoughts3 (math)	184,389	36.92%	24,380.1
safety	WildJailbreak	9,182	1.84%	1,064.8
safety	WildGuardMix	8,135	1.63%	1,343.3
safety	CoCoNot	2,173	0.44%	1,341.9
science	OpenThoughts3 (science)	25,462	5.10%	11,947.5
structured-data	TableGPT	1,145	0.23%	2,278.2