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

Split Penalty	Standard Deviation	L1 Penalty	Train LL	Validation LL	Test LL	Node#	Edge#	Feature#	Learning Time
10	0.1	2	-97.126721	-94.212348	-89.499784	864248	1981302	5670	17229.608700
10	0.1	1	-96.802491	-94.026940	-89.012877	864966	1966909	5858	18266.364090
10	0.5	2	-96.406730	-93.737268	-88.842679	851961	1948545	6092	20410.988057
10	0.1	0.1	-96.502873	-93.829723	-88.998828	852503	1967426	6090	20940.915496
10	0.5	1	-96.009291	-93.549423	-88.778822	826978	1915512	6366	26823.578195
10	0.5	0.1	-95.645793	-93.269293	-88.489396	830217	1910347	6642	32222.930369
5	0.1	2	-95.377657	-92.916819	-88.177880	849681	1947096	7452	43537.562284
5	0.1	1	-95.362718	-93.072385	-88.183673	853678	1944113	7726	57718.817405
5	0.1	0.1	-94.805957	-92.487165	-87.974580	838578	1916823	8232	67867.335596
2	0.5	0.1	-95.279330	-93.057018	-88.298937	136825	282356	9574	75209.826368
2	0.1	0.1	-94.977350	-92.636981	-87.898939	268086	583502	9948	75283.002243
2	0.1	2	-95.221413	-92.971069	-88.153827	247375	537665	9096	75416.397964
2	0.5	2	-94.579164	-92.476055	-87.824666	259431	569534	9500	75439.166502
2	0.1	1	-95.030624	-92.846275	-87.953047	237836	515176	9524	75492.812348
2	0.5	1	-95.586865	-93.328432	-88.506201	124199	254768	8958	75822.002302
5	0.5	0.1	-94.430582	-92.409842	-87.739501	577087	1310106	8744	75747.648606
5	0.5	1	-94.546585	-92.769037	-87.852913	587022	1338825	8428	76089.530632
5	0.5	2	-94.568136	-92.634079	-87.809520	648611	1475679	8252	76035.398861