microsoft_deberta-v3-base_table.md

layout	title	parent	has_children
default	microsoft_deberta-v3-base	Rankings	true

[comment]: # (This page contains a link to a table with the ranking and performance of all ranked microsoft_deberta-v3-base models. In addition, it contains a table with the baseline and the 10 best models. The original ranking was done by finetuning only the classification head of the model (linear probing) over the MNLI dataset. The best models by this ranking where ranked by the average accuracy after finetuning over the 36 datasets (except for the stsb dataset, where we used the Spearman correlation instead of accuracy).)

Ranking and performance of all 92 ranked microsoft_deberta-v3-base models (full table). The top 78 models were fully tested.

Notes:

1. The baseline results can be found here
2. While the average improvement is small, many datasets show large gains

	model_name	avg	mnli_lp	20_newsgroup	ag_news	amazon_reviews_multi	anli	boolq	cb	cola	copa	dbpedia	esnli	financial_phrasebank	imdb	isear	mnli	mrpc	multirc	poem_sentiment	qnli	qqp	rotten_tomatoes	rte	sst2	sst_5bins	stsb	trec_coarse	trec_fine	tweet_ev_emoji	tweet_ev_emotion	tweet_ev_hate	tweet_ev_irony	tweet_ev_offensive	tweet_ev_sentiment	wic	wnli	wsc	yahoo_answers
baseline	microsoft/deberta-v3-base	79.04	nan	86.41	90.44	66.86	58.78	82.99	75.00	86.57	58.40	79.43	91.93	84.48	94.49	71.86	89.78	89.20	62.26	86.73	93.51	91.79	90.42	82.35	95.06	56.98	90.28	97.76	91.02	46.19	83.95	56.21	79.82	85.06	71.80	71.21	70.21	64.09	72.03
1	sileod/deberta-v3-base-tasksource-nli	80.73	93.73	86.46	90.67	66.90	60.38	85.66	82.14	87.15	81.00	79.20	91.54	85.20	94.67	71.90	91.14	88.73	63.82	92.31	93.72	91.92	90.99	90.61	95.41	58.60	91.81	96.80	90.80	47.82	85.71	57.47	83.04	85.23	72.01	69.44	67.61	66.35	72.07
2	sileod/deberta-v3-base_tasksource-420	80.45	89.82	87.04	90.90	66.46	59.72	85.54	85.71	87.06	69.00	79.53	91.67	85.80	94.32	72.49	90.21	88.97	63.99	87.50	93.63	91.74	91.09	84.48	95.07	56.97	91.67	98.00	91.20	46.81	84.38	58.05	81.25	85.23	71.88	69.44	73.24	74.04	72.20
3	MoritzLaurer/DeBERTa-v3-base-mnli	80.37	89.77	86.42	90.63	67.18	59.34	84.16	83.93	86.29	72.00	79.20	91.42	85.10	94.21	71.51	89.88	89.95	64.03	86.54	93.67	91.69	89.87	88.09	95.41	57.38	91.38	97.40	91.60	47.30	81.28	57.64	77.17	85.35	70.29	71.79	74.65	77.88	71.70
4	mariolinml/deberta-v3-base_MNLI_10_19_v0	79.75	86.71	85.85	90.23	66.74	60.06	81.83	82.14	84.85	69.00	79.43	91.11	86.90	94.37	71.38	89.72	88.24	64.38	88.46	93.76	91.87	89.77	85.56	95.18	57.47	91.74	97.60	91.80	45.53	84.24	55.99	79.85	84.30	71.26	70.06	74.65	63.46	72.13
5	devpranjal/deberta-v3-base-devrev-data	79.58	56.73	85.49	89.93	66.72	58.19	85.26	80.36	86.29	70.00	79.03	91.18	87.70	94.01	71.25	89.48	89.95	63.63	83.65	93.81	92.03	90.24	84.12	94.72	57.65	91.44	97.20	91.40	46.37	83.25	57.68	77.42	85.93	70.93	70.85	71.83	63.46	72.50
6	nc33/deberta_finetune	79.51	75.33	86.19	90.37	67.48	58.56	84.34	73.21	86.58	68.00	79.67	91.57	88.60	94.47	72.23	89.64	90.20	63.53	87.50	93.56	91.67	90.24	83.03	95.18	58.37	90.41	97.20	90.80	47.12	85.08	59.39	79.08	83.72	70.20	70.69	67.61	64.42	72.33
7	nc33/finetune_rte_model	79.38	78.60	86.98	90.40	66.98	59.38	84.56	78.57	86.58	60.00	80.00	91.12	85.90	94.78	72.49	89.71	88.97	63.41	85.58	93.61	92.13	90.62	83.03	94.95	55.97	91.70	97.60	91.20	47.05	82.48	58.62	78.32	85.12	71.88	71.94	70.42	63.46	72.13
8	MoritzLaurer/DeBERTa-v3-base-mnli-fever-anli	79.37	89.93	86.42	90.40	67.42	58.47	84.50	92.86	85.43	50.00	79.40	91.44	78.60	94.36	71.90	89.39	89.95	64.11	85.58	93.81	92.01	90.15	88.45	95.30	55.48	91.81	97.40	91.60	45.25	82.76	54.95	77.30	85.58	71.53	68.81	73.24	70.19	71.40
9	s8n29/finetuned_model_1	79.25	57.45	86.06	90.43	66.52	57.78	85.08	76.79	86.86	72.00	78.97	90.91	83.50	94.06	72.95	89.70	89.22	63.26	85.58	94.14	92.05	90.34	79.78	94.84	57.01	91.05	97.00	91.20	45.48	83.18	57.71	80.74	85.12	69.96	70.06	67.61	63.46	72.57
10	koolerkx/autotrain-sns-fake-news-3229590413	79.24	65.50	86.78	91.07	66.98	60.44	85.14	75.00	86.19	52.00	79.43	90.90	86.30	94.39	72.29	89.36	89.22	62.48	89.42	93.68	92.04	90.71	81.23	94.84	58.73	91.20	96.80	91.80	46.93	83.67	54.78	79.08	85.00	71.61	71.94	76.06	63.46	71.83

Download full models ranking table: [csv](./results/microsoft_deberta-v3-base_table.csv)

Home