Complete result settings for Reuters-52 dataset (table columns are sortable)
The experiments are described in Rooshenas and Lowd, Discriminative Structure Learning of Arithmetic Circuits, AIStats 16

Tree#	C	Max. EM Iteration	Train CLL	Validation CLL	Test CLL	Learning Time
1	0.100000	10	-12.069705	-17.137962	-17.251693	16319.430070
1	2.000000	10	-5.156270	-22.170601	-22.279998	18738.136543
1	2.000000	20	-5.156270	-22.170601	-22.279998	19371.600633
1	0.100000	20	-12.069705	-17.137962	-17.251693	19446.141666
1	0.100000	50	-12.069705	-17.137962	-17.251693	19933.081871
1	0.500000	10	-7.643366	-18.530966	-18.587359	20000.853200
1	1.000000	20	-6.321181	-20.164755	-20.216109	20058.587360
1	2.000000	50	-5.156270	-22.170601	-22.279998	20290.065210
1	0.500000	20	-7.643366	-18.530966	-18.587359	21432.072050
1	0.500000	50	-7.643366	-18.530966	-18.587359	22994.113333
1	1.000000	10	-6.321181	-20.164755	-20.216109	23383.395051
1	1.000000	50	-6.321181	-20.164755	-20.216109	34554.806644
2	0.100000	10	-9.596552	-17.521472	-17.581108	71012.069000
2	0.500000	10	-5.344999	-19.472900	-19.393563	72324.212753
2	0.100000	20	-9.596552	-17.521472	-17.581108	72914.770116
2	0.500000	20	-5.344999	-19.472900	-19.393563	73699.811112
2	0.500000	50	-5.344999	-19.472900	-19.393563	74007.660367
2	1.000000	20	-4.126168	-21.233582	-21.178172	77523.137128
2	0.100000	50	-9.596552	-17.521472	-17.581108	78719.220326
2	2.000000	10	-3.116596	-22.990839	-23.289103	79014.929272
2	1.000000	10	-4.126168	-21.233582	-21.178172	83971.676139
1	0.010000	10	-20.740455	-21.238629	-21.542158	17489.995965
1	0.050000	10	-14.373071	-17.524949	-17.707225	18627.586945