Cross-lingual Retrieval-Augmented Generation on BioASQ-ENKB5 benchmark leaderboard · SOTA2 Research

Cross-lingual Retrieval-Augmented Generation on BioASQ-ENKB5

90.99ID ScoreTR-RAG

Evaluation Results

Method	Links
TR-RAG2026.07		90.99	97.42	—	—	—	97.03	99.15	98.41	96.4	99.68	99.25	99.47	97.42
Llama-3.3-70B-Instruct2026.07		90.78	97.68	—	—	—	98.51	99.25	98.41	96.74	99.78	99.36	99.57	97.68
Llama-3.2-3B-Instruct2026.07		87.71	96.34	—	—	—	95.76	98.72	97.66	94.96	99.78	98.41	99.1	96.34
TR-RAG2026.07		81.03	—	—	—	74.84	64.34	71.4	80.77	74.38	86.63	64.89	75.76	74.84
TR-RAG2026.07		80.16	—	58.09	—	—	29.55	48.69	73.54	57.99	85.14	31.46	58.3	58.09
Naive-RL2026.07		78.98	—	—	—	73.06	63.13	69	78.77	72.47	84.65	63.8	74.23	73.06
On-Policy Distillation2026.07		74.25	—	—	—	70.01	60.94	64.97	75.03	68.8	82.59	62.25	72.42	70.01
SDFT2026.07		73.24	—	—	—	69.16	60.21	64.07	73.92	67.86	82.07	61.46	71.77	69.16
Knowledge distillation2026.07		72.67	—	—	—	68.7	59.02	63.84	73.28	67.2	81.88	61.52	71.7	68.7
TR-RAG2026.07		71.94	—	—	69.02	—	66.44	66.36	70.35	68.77	75.08	63.97	69.53	69.02
DIT2026.07		71.34	—	—	—	65.61	52.84	58.88	71.51	63.64	81.16	57.93	69.55	65.61
Prompt-Control2026.07		71.05	—	—	—	67	55.65	61.71	72.83	65.31	81.96	58.82	70.39	67
Translate2026.07		70.81	—	—	—	63	48.76	55.7	70.03	61.33	80.98	51.72	66.35	63
SFT2026.07		62.49	—	—	—	63.14	57.24	60.44	64.16	61.08	74.03	60.5	67.27	63.14