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

499,449 samples from lightonai/Dolci-Think-SFT-32B-500k-FR

Category Distribution

Category	Count	%	Mean Tokens	Median	Std	Min	Max	Total Tokens
math	184,444	36.93%	24,306.2	23,900	7,344.0	1,371	45,452	4,483,128,648
code	169,822	34.00%	9,246.8	5,763	8,721.9	371	44,199	1,570,310,174
instruction-following	80,952	16.21%	3,027.0	1,698	3,811.1	38	41,035	245,043,357
science	25,458	5.10%	11,811.3	10,518	7,260.4	972	40,705	300,693,121
safety	19,490	3.90%	1,175.5	1,072	553.0	144	16,045	22,911,076
general-chat	18,138	3.63%	4,195.8	2,884	3,903.1	57	47,880	76,103,349
structured-data	1,145	0.23%	2,265.9	1,686	1,958.4	260	15,344	2,594,488

Token Count Statistics

Stat	Value
Count	499,449
Mean	13,416.4
Std	11,311.5
Min	38
25%	2,462
Median	10,844
75%	22,137
90%	30,595
95%	33,648
99%	37,365
Max	47,880
Total	6,700,784,213

Dataset to Category Mapping

Category	Original Dataset	Count	%	Mean Tokens
code	Dolci-Think-Python	121,918	24.41%	5,688.8
code	Nemotron Post-Training	27,325	5.47%	15,378.0
code	OpenThoughts3 (code)	20,579	4.12%	22,185.0
general-chat	WildChat (Tulu v3.9)	9,739	1.95%	4,296.3
general-chat	WildChat	7,600	1.52%	4,281.0
general-chat	OpenAssistant (OASST1)	799	0.16%	2,160.3
instruction-following	Dolci-Precise-IF	53,699	10.75%	2,054.0
instruction-following	SYNTHETIC-2 (IF/QwQ)	27,253	5.46%	4,944.3
math	OpenThoughts3 (math)	184,444	36.93%	24,306.2
safety	WildJailbreak	9,182	1.84%	1,036.4
safety	WildGuardMix	8,135	1.63%	1,291.8
safety	CoCoNot	2,173	0.44%	1,328.0
science	OpenThoughts3 (science)	25,458	5.10%	11,811.3
structured-data	TableGPT	1,145	0.23%	2,265.9