Share your thoughts, 1 month free Claude Pro on usSee more

Retrieval-Augmented Generation on Hotpot ENKB5 (In-Domain Average)

86.51Language Consistency

TR-RAG

Updated 18d ago

Evaluation Results

Method	Links
TR-RAG 2026.07		86.51	67.96	84.67	79.68	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
SFT 2026.07		83.77	54.66	71.32	69.92	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
Llama-3.2-3B-Instruct 2026.07		81.93	33.09	53.16	56.06	56.72	99.63	99.27	72.09	44.99	18.81	24.18	44.38	54.87	49.5	50.6	57.65	52.19	55.98	58.02	58.04
Knowledge distillation 2026.07		81.57	46.12	67.68	65.12	55.55	99.72	99.81	71.18	63.04	27.72	34.63	59.1	72.4	60.79	64.29	73.22	63.66	62.74	66.24	67.83
Prompt-Control 2026.07		81.43	36.3	52.96	56.9	55.09	99.63	99.81	71.18	48.67	20.73	26.2	49.6	56.93	46.84	49.11	58.96	53.56	55.73	58.37	59.91
DIT 2026.07		81.43	34.99	46.68	54.36	55.81	99.36	99.81	70.72	50.37	17.83	19.85	51.89	52.53	43.15	35.4	55.62	52.9	53.45	51.69	59.41
TR-RAG 2026.07		81.41	65.35	83.67	76.81	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
SDFT 2026.07		81.34	49.34	72.32	67.67	55	99.54	99.72	71.09	67.25	29.87	35.37	64.87	75.02	65.99	70.53	77.74	65.76	65.13	68.54	71.23
On-Policy Distillation 2026.07		81.32	50.76	72.49	68.19	54.73	99.64	99.82	71.09	70.01	31.98	35.61	65.42	76.31	66.1	69.41	78.15	67.02	65.91	68.28	71.55
Naive-RL 2026.07		81.3	52.37	75.59	69.75	54.64	99.73	99.82	71	69.87	29.58	43.15	66.87	79.72	69.8	72.38	80.44	68.08	66.37	71.78	72.77
TR-RAG 2026.07		81.29	54.96	76.14	70.8	54.45	99.72	99.81	71.18	72.2	33.77	45.37	68.5	79.66	69.74	73.65	81.52	68.77	67.74	72.94	73.73
Llama-3.3-70B-Instruct 2026.07		80.93	59.73	82.86	74.51	55	99.22	99.09	70.4	78.89	44.26	44.57	71.21	83.74	81.25	81.31	85.13	72.54	74.91	74.99	75.58
Translate 2026.07		80.7	26.21	32.46	46.46	55.72	97.27	98.91	70.91	40.06	11.37	18.28	35.13	44.83	19.93	25.47	39.62	46.87	42.86	47.55	48.55
SFT 2026.07		79.87	46.67	62.04	62.86	63.73	90	95.18	70.55	57.21	34.02	39.31	56.14	69.03	52.35	57.19	69.57	63.32	58.79	63.89	65.42
Naive-RL 2026.07		73.09	66.96	84.3	74.79	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-