Space Fake

Fake

Data (20k): https://www.kaggle.com/c/fake-news/data?select=train.csv

Model	Epochs	Train Params	loss	accuracy	cs accuracy	f1	cs f1	precision	recall
DistilBERT	0?	592130	0.6182	0.7478	N/A	0.7372	N/A	0.8549	0.5707
Space-DistilBERT	0?	197122 (128)	0.5986	0.7407	0.	0.7281	0.	0.8602	0.5481
Space-DistilBERT	0?	394242 (256)	0.5714	0.7655	0.	0.7578	0.	0.8566	0.6133

Fake 2 (not used since mainly a subset of Fake 1)

Data (6k): https://www.kaggle.com/datasets/rajatkumar30/fake-news

Model	Train Params	loss	accuracy	cs accuracy	f1	cs f1	precision	recall
DistilBERT	592130	0.6453	0.7572	N/A	0.7567	N/A	0.7848	0.7106
Space-DistilBERT	197122 (128)	0.6356	0.7749	0.	0.7691	0.	0.9069	0.6142
Space-DistilBERT	394242 (256)	0.6113	0.7976	0.	0.7942	0.	0.9040	0.6673

Zero-Shot

(train on 20k)

Model	Train Params	loss	accuracy	f1	precision	recall
DistilBERT	592130	0.6287	0.7704	0.7704	0.7820	0.7494
Space-DistilBERT	197122 (128)	0.6175	0.7313	0.7231	0.8521	0.5591
Space-DistilBERT	394242 (256)	0.6014	0.7501	0.7467	0.8246	0.6346

Covid Fake News Data

Model	Epochs	Train Params	loss	accuracy	cs accuracy	f1	cs f1	precision	recall	Inter-space weight	Intra-space weight
DistilBERT	5	592130	0.6287	0.8107	N/A	0.8027	N/A	0.9462	0.6392	N/A	N/A
DistilBERT	750/1800	592130	0.1572	0.9313	N/A	0.9310	N/A	0.9431	0.9108	N/A	N/A
Space-DistilBERT	1450/1800	6162 (4)	0.3134	0.9285	0.8706	0.9283	0.8690	0.9348	0.9137	0.1	1e-5
Space-DistilBERT	1450/1800	197122 (128)	0.3645	0.9304	0.7850	0.9302	0.7799	0.9333	0.9196	0.1	1e-5
Space-DistilBERT	5	394242 (256)	0.5810	0.7962	0.	0.7867	0.	0.9371	0.6137	0	0
Space-DistilBERT	5	788482 (512)	0.5550	0.7888	0.	0.7834	0.	0.8622	0.6627	0	0
Space-DistilBERT	5	788482 (512)	0.7514	0.7929	0.	0.7875	0.	0.8712	0.6637	0.1	0
Space-BERT	70	197122 (128)	0.2498	0.8957	0.	0.8950	0.	0.9217	0.8539	0	0
Space-BERT	70	6162 (4)	0.5506	0.7991	0.	0.7935	0.	0.8831	0.6667	0	0
Space-BERT	500	4622 (3)	0.4145	0.8995	0.8860	0.8988	0.8844	0.9305	0.8529	0.1	1e-5
Space-BERT	500	4622 (3)	0.2780	0.9051	0.4453	0.9045	0.4438	0.9295	0.8667	0	0
Space-BERT	250	98562 (64)	0.3907	0.9182	0.7976	0.9179	0.7905	0.9316	0.8941	0.1	1e-5
BERT	250	1538	0.5175	0.7897	N/A	0.7861	N/A	0.8376	0.6931	N/A	N/A
BERT	5	108311810	0.0846	0.9752	N/A	0.9752	N/A	0.9736	0.9745	N/A	N/A
BERT	50	1538	0.6338	0.7145	N/A	0.6966	N/A	0.7528	0.7047	N/A	N/A
Space-BERT	50	4622 (3)	0.7875	0.7949	0.5234	0.7828	0.3436	0.8478	0.7855	0.1	1e-5
Space-BERT	50	98562 (64)	0.5877	0.8645	0.7743	0.8627	0.7637	0.8725	0.8610	0.1	1e-5
Space-BERT	50	197122 (128)	0.5727	0.8808	0.6528	0.8797	0.5967	0.8859	0.8782	0.1	1e-5

Notes: for the journal version we want to prove that either we are better with short training and eventually, or we are better eventually. We need to present the results for both cases. For the small fake news dataset, we need to show that we are better with at least one of the two training.

Zero-Shot with CS Loss

(test on 6k) (this is a data leakage) these results are cross-dataset without any specific topic

Model	Epochs	Train Params	loss	accuracy	cs accuracy	f1	cs f1	precision	recall	Inter-space weight	Intra-space weight
BERT	50	1538	0.6544	0.6230	N/A	0.6028	N/A	0.7227	0.3979	N/A	N/A
Space-BERT	50	4622 (3)	0.5722	0.7679	0.5958	0.7657	0.5902	0.8322	0.6707	0.1	1e-5
Space-BERT*	50	98562 (64)	0.4534	0.7975	0.6148	0.7974	0.5809	0.8040	0.7860	0.1	1e-5
Space-BERT	50	197122 (128)	0.4567	0.7964	0.5863	0.7963	0.5299	0.7822	0.8208	0.1	1e-5

this is after intersection removal:

Model	Epochs	Train Params	loss	accuracy	cs accuracy	f1	cs f1	precision	recall	Inter-space weight	Intra-space weight
BERT	50	1538	0.6035	0.7736	N/A	0.6044	N/A	0.5985	0.6132	N/A	N/A
Space-BERT	50	4622 (3)	0.5544	0.8292	0.5154	0.7180	0.4711	0.6985	0.7515	0.1	1e-5
Space-BERT	50	98562 (64)	0.4666	0.7997	0.8071	0.7069	0.6153	0.6847	0.7798	0.1	1e-5
Space-BERT	50	197122 (128)	0.5079	0.7750	0.8201	0.6892	0.5904	0.6729	0.7816	0.1	1e-5

* - our assumption was right, the model with less parameters but higher CS scores is better in terms of generalization

(test on covid-fake) these results are cross-dataset with specific (covid) topic

Model	Epochs	Train Params	loss	accuracy	cs accuracy	f1	cs f1	precision	recall	Inter-space weight	Intra-space weight
BERT	50	1538	0.7692	0.4763	N/A	0.3308	N/A	0.4762	0.9889	N/A	N/A
Space-BERT	50	4622 (3)	0.9535	0.4716	0.4766	0.3215	0.3233	0.4740	0.9882	0.1	1e-5
Space-BERT	50	98562 (64)	2.1427	0.4597	0.4709	0.3169	0.3214	0.4675	0.9618	0.1	1e-5
Space-BERT	50	197122 (128)	2.5363	0.4530	0.4092	0.3149	0.3116	0.4638	0.9461	0.1	1e-5

Cross-dataset benchmarking

Dataset	Total true news	Total fake news	Images used	Public availability
FakeCovid (twitter)	3360	3060	No	Yes
LIAR (multi-label)	6400	6400	No	Yes
LIAR (binary-label)	6400	6400	No	Yes
Fake News Kaggle Competition	10387	10413	No	Yes
FakeNewsNet	18,000	6,000	Yes	Yes

Kaggle Fake News Competition

Model	Epochs	Train Params	loss	accuracy	cs accuracy	f1	cs f1	precision	recall	Inter-space weight	Intra-space weight
BERT	50	1538	0.6251	0.6550	N/A	0.6337	N/A	0.7400	0.4319	N/A	N/A
Space-BERT	50	4622 (3)	0.7444	0.8069	0.6408	0.8030	0.6396	0.8769	0.6946	0.1	1e-5
Space-BERT	50	98562 (64)	0.5305	0.8685	0.6834	0.8672	0.6497	0.9132	0.8018	0.1	1e-5
Space-BERT	50	197122 (128)	0.5016	0.8868	0.6436	0.8859	0.5900	0.9228	0.8334	0.1	1e-5

LIAR (multi-label)

Model	Epochs	Train Params	loss	accuracy	cs accuracy	f1	cs f1	precision	recall	Inter-space weight	Intra-space weight
BERT	50	4614	1.7426	0.2221	N/A	0.1211	N/A	0.1261	0.1815	N/A	N/A
Space-BERT	50	13938 (3)	2.2877	0.2362	0.1824	0.1540	0.1079	0.1542	0.1977	0.1	1e-5
Space-BERT	50	297222 (64)	2.3192	0.2580	0.2362	0.2034	0.1590	0.2201	0.2227	0.1	1e-5
Space-BERT	50	594438 (128)	2.3651	0.2572	0.2081	0.2120	0.1267	0.2527	0.2248	0.1	1e-5

LIAR (binary-label)

Model	Epochs	Train Params	loss	accuracy	cs accuracy	f1	cs f1	precision	recall	Inter-space weight	Intra-space weight
BERT	50	1538	0.7026	0.5666	N/A	0.3617	N/A	0.2833	0.5000	N/A	N/A
Space-BERT	50	4622 (3)	0.8476	0.5900	0.5838	0.5280	0.5825	0.5778	0.5515	0.1	1e-5
Space-BERT	50	98562 (64)	0.8591	0.6267	0.5877	0.6002	0.5855	0.6185	0.6031	0.1	1e-5
Space-BERT	50	197122 (128)	0.8747	0.6251	0.5744	0.5971	0.4194	0.6170	0.6007	0.1	1e-5

FakeNewsNet (GossipCop)

Model	Epochs	Train Params	loss	accuracy	cs accuracy	f1	cs f1	precision	recall	Inter-space weight	Intra-space weight
BERT	50	1538	0.5480	0.7566	N/A	0.4307	N/A	0.3783	0.5000	N/A	N/A
Space-BERT	50	4622 (3)	0.7111	0.7575	0.2620	0.4344	0.2238	0.8786	0.5017	0.1	1e-5
Space-BERT	50	98562 (64)	0.6771	0.7933	0.6378	0.6620	0.6037	0.7275	0.6427	0.1	1e-5
Space-BERT	50	197122 (128)	0.6847	0.7987	0.7143	0.6776	0.6361	0.7353	0.6573	0.1	1e-5

FakeNewsNet (PolitiFact + GossipCop)

Model	Epochs	Train Params	loss	accuracy	cs accuracy	f1	cs f1	precision	recall	Inter-space weight	Intra-space weight
BERT	50	1538	0.5502	0.7548	N/A	0.4302	N/A	0.3774	0.5000	N/A	N/A
Space-BERT	50	4622 (3)	0.7124	0.7557	0.2732	0.4336	0.2400	0.8777	0.5016	0.1	1e-5
Space-BERT	50	98562 (64)	0.6754	0.7955	0.6468	0.6654	0.6090	0.7359	0.6453	0.1	1e-5
Space-BERT	50	197122 (128)	0.6826	0.8028	0.7131	0.6833	0.6302	0.7475	0.6616	0.1	1e-5

FakeNewsNet (GossipCop predict> Covid Fake)

Model	Epochs	Train Params	loss	accuracy	cs accuracy	f1	cs f1	precision	recall	Inter-space weight	Intra-space weight
BERT	50	1538	0.8384	0.5234	N/A	0.3436	N/A	0.2616	0.5000	N/A	N/A
Space-BERT	50	4622 (3)	0.7862	0.5234	0.6145	0.3436	0.6052	0.2617	0.5000	0.1	1e-5
Space-BERT	50	98562 (64)	0.8823	0.5375	0.6064	0.3806	0.5874	0.6911	0.5151	0.1	1e-5
Space-BERT	50	197122 (128)	0.9444	0.5373	0.5691	0.3797	0.4712	0.6954	0.5149	0.1	1e-5

FakeNewsNet (GossipCop predict> PolitiFact)

Model	Epochs	Train Params	loss	accuracy	cs accuracy	f1	cs f1	precision	recall	Inter-space weight	Intra-space weight
BERT	50	1538	0.5990	0.5909	N/A	0.3714	N/A	0.2955	0.5000	N/A	N/A
Space-BERT	50	4622 (3)	0.5621	0.5909	0.4403	0.3714	0.3699	0.2955	0.5000	0.1	1e-5
Space-BERT	50	98562 (64)	0.5385	0.6695	0.6259	0.6085	0.6258	0.6812	0.6181	0.1	1e-5
Space-BERT	50	197122 (128)	0.5312	0.6941	0.6468	0.6395	0.6155	0.7172	0.6447	0.1	1e-5

FakeNewsNet (PolitiFact + GossipCop predict> Covid Fake)

Model	Epochs	Train Params	loss	accuracy	cs accuracy	f1	cs f1	precision	recall	Inter-space weight	Intra-space weight
BERT	50	1538	0.8171	0.5234	N/A	0.3436	N/A	0.2617	0.5000	N/A	N/A
Space-BERT	50	4622 (3)	0.7652	0.5234	0.6024	0.3436	0.5881	0.2617	0.5000	0.1	1e-5
Space-BERT	50	98562 (64)	0.8256	0.5482	0.6191	0.4064	0.6116	0.6983	0.5265	0.1	1e-5
Space-BERT	50	197122 (128)	0.8719	0.5480	0.5956	0.4056	0.5354	0.7008	0.5263	0.1	1e-5

Ablation Study with inter-space and intra-space losses only

FakeNewsNet (PolitiFact + GossipCop)

Model	Epochs	Train Params	loss	accuracy	cs accuracy	f1	cs f1	precision	recall	Inter-space weight	Intra-space weight
Space-BERT	10	4622 (3)	1.7721	0.7548	0.6196	0.4302	0.5962	0.3774	0.5000	1.0	1e-5
Space-BERT	10	98562 (64)	0.7079	0.7815	0.7055	0.6016	0.6325	0.7203	0.5928	1.0	1e-5
Space-BERT	10	197122 (128)	0.6910	0.7973	0.7321	0.6703	0.6370	0.7350	0.6500	1.0	1e-5
Space-BERT	10	4622 (3)	1.9239	0.7548	0.7152	0.4301	0.6437	0.3774	0.5000	1.0	0
Space-BERT	10	98562 (64)	1.9734	0.2565	0.6616	0.2145	0.6254	0.5756	0.5056	1.0	0
Space-BERT	10	197122 (128)	1.9751	0.7349	0.7637	0.4343	0.6567	0.4329	0.4909	1.0	0

FakeNewsNet (GossipCop)

Model	Epochs	Train Params	loss	accuracy	cs accuracy	f1	cs f1	precision	recall	Inter-space weight	Intra-space weight
Space-BERT	10	4622 (3)	1.9304	0.7566	0.7053	0.4307	0.6387	0.3783	0.5000	1.0	1e-5
Space-BERT	10	98562 (64)	2.0755	0.2818	0.6857	0.2538	0.6366	0.5683	0.5175	1.0	1e-5
Space-BERT	10	197122 (128)	2.0751	0.7448	0.7442	0.4436	0.6522	0.4861	0.4985	1.0	1e-5
Space-BERT	10	4622 (3)	1.9255	0.7566	0.7055	0.4307	0.6387	0.3783	0.5000	1.0	0
Space-BERT	10	98562 (64)	1.9739	0.2549	0.6761	0.2137	0.6361	0.5737	0.5057	1.0	0
Space-BERT	10	197122 (128)	1.9754	0.7369	0.7651	0.4335	0.6659	0.4277	0.4905	1.0	0

FakeNewsNet (PolitiFact + GossipCop predict> Covid Fake)

Model	Epochs	Train Params	loss	accuracy	cs accuracy	f1	cs f1	precision	recall	Inter-space weight	Intra-space weight
Space-BERT	10	4622 (3)	0.7046	0.5234	0.5379	0.3436	0.3830	0.2617	0.5000	1.0	1e-5
Space-BERT	10	98562 (64)	0.6888	0.4761	0.6260	0.3238	0.6039	0.4200	0.4993	1.0	1e-5
Space-BERT	10	197122 (128)	0.7055	0.4338	0.5778	0.3467	0.4927	0.3457	0.4177	1.0	1e-5
Space-BERT	10	4622 (3)	0.7034	0.5234	0.5384	0.3436	0.3849	0.2617	0.5000	1.0	0
Space-BERT	10	98562 (64)	0.6915	0.4762	0.6305	0.3229	0.6049	0.3493	0.4995	1.0	0
Space-BERT	10	197122 (128)	0.7039	0.4349	0.5969	0.3252	0.5276	0.3042	0.4169	1.0	0

FakeNewsNet (GossipCop predict> PolitiFact)

Model	Epochs	Train Params	loss	accuracy	cs accuracy	f1	cs f1	precision	recall	Inter-space weight	Intra-space weight
Space-BERT	10	4622 (3)	0.7034	0.5234	0.5292	0.3436	0.3606	0.2617	0.5000	1.0	1e-5
Space-BERT	10	98562 (64)	0.6892	0.4759	0.5967	0.3238	0.5549	0.4112	0.4991	1.0	1e-5
Space-BERT	10	197122 (128)	0.7052	0.4401	0.5543	0.3472	0.4312	0.3468	0.4234	1.0	1e-5
Space-BERT	10	4622 (3)	0.6918	0.4762	0.5947	0.3229	0.5399	0.3493	0.4995	1.0	0
Space-BERT	10	98562 (64)	0.7022	0.5234	0.5289	0.3436	0.3598	0.2617	0.5000	1.0	0
Space-BERT	10	197122 (128)	0.7037	0.4546	0.5653	0.3290	0.4510	0.3046	0.4353	1.0	0

Explanation Benchmark

---

Explanation:

• Train the model on some dataset (e.g. Fake News Kaggle Competition, or Covid Fake News Dataset)
• Let's embed the whole fact-checking dataset into the same space (not the embeddings, but the centroid of the embeddings).
• Let's try to predict the news to be fake or true.
• After we've predicted let's use the embedding centroid and extract k nearest neighbors from the knowledge base.
• Let's see how well this nearest neighbors match the original fact-checking articles (this we will measure by max/average cosine/euclidian similarity of the neighbors embedding with the original fact-checking articles, and by the number of exact matches, e.g. recall and precision).

Notes: we should actually compare this with S-BERT. For now we just use Mean of BERT embeddings as a centroid. We should also think how we can deal with the fact that to use concept space similarity we need to use N knowledge bases - one for each label. Since otherwise vectors that fall in different spaces will be compared (since we force them to be orthogonal only with those from the different concept spaces).

---

tp / (tp + fp) = precision (how well we identify true explanation) tp / (tp + fn) = recall (how well we distinguish true explanation from false explanation)

Model	Train Dataset	Epochs	Train Params	Mean Cosine Similarity	Max Cosine Similarity	Mean Euclidean Distance	Max Euclidean Distance	precision	recall
BERT ✅	Covid Fake	50	1538	0.	0.	0.	0.	0.	0.
BERT ✅	Fake News Kaggle	50	1538	0.	0.	0.	0.	0.	0.
Space-BERT ✅	Covid Fake	50	4622 (3)	0.	0.	0.	0.	0.	0.
Space-BERT ✅	Covid Fake	50	98562 (64)	0.	0.	0.	0.	0.	0.
Space-BERT ✅	Covid Fake	50	197122 (128)	0.	0.	0.	0.	0.	0.
Space-BERT ✅	Fake News Kaggle	50	4622 (3)	0.	0.	0.	0.	0.	0.
Space-BERT ✅	Fake News Kaggle	50	98562 (64)	0.	0.	0.	0.	0.	0.
Space-BERT ✅	Fake News Kaggle	50	197122 (128)	0.	0.	0.	0.	0.	0.

Some tests that are used to make sure that model works as expected

IMDB

Experiment	Epochs	Train Params	Loss	Accuracy	F1-score (macro)	Precision	Recall	Inter-space weight	Intra-space weight
Space-DistilBERT (CE + inter-space loss)	5	4622	0.8804	0.6141	0.5587	0.8957	0.2594	0	0
Space-DistilBERT (CE loss)	5	4622	0.4883	0.8080	0.8079	0.8262	0.7808	0	0
Space-DistilBERT (CE loss)	5	197122	0.3855	0.8322	0.8320	0.8093	0.8663	0	0
Space-DistilBERT (CE + inter-space loss)	5	197122	0.7847	0.7890	0.7889	0.8016	0.7687	0.1	0
DistilBERT-base-cased	5	592130	0.4612	0.7852	0.7819	0.8799	0.6614	N/A	N/A

Paper tables:

Same dataset benchmarking

Data	Model	# Latent Dimensions	loss	accuracy	cs accuracy	f1	cs f1	precision	recall
Fake COVID News	BERT	N/A	0.6338	0.7145	N/A	0.6966	N/A	0.7528	0.7047
	Space-BERT	3	0.7875	0.7949	0.5234	0.7828	0.3436	0.8478	0.7855
	Space-BERT	64	0.5877	0.8645	0.7743	0.8627	0.7637	0.8725	0.8610
	Space-BERT	128	0.5727	0.8808	0.6528	0.8797	0.5967	0.8859	0.8782
Liar (multi-label)	BERT	N/A	1.7426	0.2221	N/A	0.1211	N/A	0.1261	0.1815
	Space-BERT	3	2.2877	0.2362	0.1824	0.1540	0.1079	0.1542	0.1977
	Space-BERT	64	2.3192	0.2580	0.2362	0.2034	0.1590	0.2201	0.2227
	Space-BERT	128	2.3651	0.2572	0.2081	0.2120	0.1267	0.2527	0.2248
Liar (binary-label)	BERT	N/A	0.7026	0.5666	N/A	0.3617	N/A	0.2833	0.5000
	Space-BERT	3	0.8476	0.5900	0.5838	0.5280	0.5825	0.5778	0.5515
	Space-BERT	64	0.8591	0.6267	0.5877	0.6002	0.5855	0.6185	0.6031
	Space-BERT	128	0.8747	0.6251	0.5744	0.5971	0.4194	0.6170	0.6007
Kaggle Fake News	BERT	N/A	0.6251	0.6550	N/A	0.6337	N/A	0.7400	0.4319
	Space-BERT	3	0.7444	0.8069	0.6408	0.8030	0.6396	0.8769	0.6946
	Space-BERT	64	0.5305	0.8685	0.6834	0.8672	0.6497	0.9132	0.8018
	Space-BERT	128	0.5016	0.8868	0.6436	0.8859	0.5900	0.9228	0.8334
Fake News Net	BERT	N/A	0.5502	0.7548	N/A	0.4302	N/A	0.3774	0.5000
	Space-BERT	3	0.7124	0.7557	0.2732	0.4336	0.2400	0.8777	0.5016
	Space-BERT	64	0.6754	0.7955	0.6468	0.6654	0.6090	0.7359	0.6453
	Space-BERT	128	0.6826	0.8028	0.7131	0.6833	0.6302	0.7475	0.6616

Cross-dataset benchmarking

(Train) -> (Test)	Model	# Dims.	loss	accuracy	cs accuracy	f1	cs f1	precision	recall
(GossipCop) -> (CovidFake)	BERT	N/A	0.8384	0.5234	N/A	0.3436	N/A	0.2616	0.5000
	Space-BERT	3	0.7862	0.5234	0.6145	0.3436	0.6052	0.2617	0.5000
	Space-BERT	64	0.8823	0.5375	0.6064	0.3806	0.5874	0.6911	0.5151
	Space-BERT	128	0.9444	0.5373	0.5691	0.3797	0.4712	0.6954	0.5149
(GossipCop) -> (Politifact)	BERT	N/A	0.5990	0.5909	N/A	0.3714	N/A	0.2955	0.5000
	Space-BERT	3	0.5621	0.5909	0.4403	0.3714	0.3699	0.2955	0.5000
	Space-BERT	64	0.5385	0.6695	0.6259	0.6085	0.6258	0.6812	0.6181
	Space-BERT	128	0.5312	0.6941	0.6468	0.6395	0.6155	0.7172	0.6447
(FakeNewsNet) -> (CovidFake)	BERT	N/A	0.8171	0.5234	N/A	0.3436	N/A	0.2617	0.5000
	Space-BERT	3	0.7652	0.5234	0.6024	0.3436	0.5881	0.2617	0.5000
	Space-BERT	64	0.8256	0.5482	0.6191	0.4064	0.6116	0.6983	0.5265
	Space-BERT	128	0.8719	0.5480	0.5956	0.4056	0.5354	0.7008	0.5263

(Train) -> (Test)	Model	# Dims.	loss	accuracy	cs accuracy	f1	cs f1
(Gossip) -> (CovidFake)	BERT	N/A	0.8384	0.5234	N/A	0.3436	N/A
	Space-BERT	3	0.7862	0.5234	0.6145	0.3436	0.6052
	Space-BERT	64	0.8823	0.5375	0.6064	0.3806	0.5874
	Space-BERT	128	0.9444	0.5373	0.5691	0.3797	0.4712
(Gossip) -> (Politifact)	BERT	N/A	0.5990	0.5909	N/A	0.3714	N/A
	Space-BERT	3	0.5621	0.5909	0.4403	0.3714	0.3699
	Space-BERT	64	0.5385	0.6695	0.6259	0.6085	0.6258
	Space-BERT	128	0.5312	0.6941	0.6468	0.6395	0.6155
(NewsNet) -> (CovidFake)	BERT	N/A	0.8171	0.5234	N/A	0.3436	N/A
	Space-BERT	3	0.7652	0.5234	0.6024	0.3436	0.5881
	Space-BERT	64	0.8256	0.5482	0.6191	0.4064	0.6116
	Space-BERT	128	0.8719	0.5480	0.5956	0.4056	0.5354

Ablation Study

Same dataset inter-space and intra-space loss ablation study

Data	Model	# Latent Dimensions	loss	accuracy	cs accuracy	f1	cs f1	precision	recall
GossipCop	Space-BERT	3	1.9304	0.7566	0.7053	0.4307	0.6387	0.3783	0.5000
	Space-BERT	64	2.0755	0.2818	0.6857	0.2538	0.6366	0.5683	0.5175
	Space-BERT	128	2.0751	0.7448	0.7442	0.4436	0.6522	0.4861	0.4985
Fake News Net	Space-BERT	3	1.7721	0.7548	0.6196	0.4302	0.5962	0.3774	0.5000
	Space-BERT	64	0.7079	0.7815	0.7055	0.6016	0.6325	0.7203	0.5928
	Space-BERT	128	0.6910	0.7973	0.7321	0.6703	0.6370	0.7350	0.6500

Cross-dataset inter-space and intra-space loss ablation study

(Train) -> (Test)	Model	# Dims.	loss	accuracy	cs accuracy	f1	cs f1	precision	recall
(GossipCop) -> (CovidFake)	Space-BERT	3	0.7034	0.5234	0.5292	0.3436	0.3606	0.2617	0.5000
	Space-BERT	64	0.6892	0.4759	0.5967	0.3238	0.5549	0.4112	0.4991
	Space-BERT	128	0.7052	0.4401	0.5543	0.3472	0.4312	0.3468	0.4234
(FakeNewsNet) -> (CovidFake)	Space-BERT	3	0.7046	0.5234	0.5379	0.3436	0.3830	0.2617	0.5000
	Space-BERT	64	0.6888	0.4761	0.6260	0.3238	0.6039	0.4200	0.4993
	Space-BERT	128	0.7055	0.4338	0.5778	0.3467	0.4927	0.3457	0.4177