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

Question Answering on NQ (Natural Questions) (EM)

78.3EM

InstructRetro 43B

Updated 1mo ago

Evaluation Results

Method	Links
InstructRetro 43B 2024.07		78.3
Raven 11B 2024.07		65.7
Llama3-RankRAG 70B 2024.07		54.2
DeepSeek-V2 236B 2024.07		53.4
Llama3-RankRAG 8B 2024.07		50.6
BAR-RAG 2026.02		49.5
BAR-RAG 2026.02		49
BAR-RAG 2026.02		47.5
Llama3-ChatQA-1.5 70B 2024.07		47
BAR-RAG 2026.02		46.9
GPT-3.5-turbo-1106 RAG 2024.07		46.7
SubSearch-base (Ours) 2026.04		46.3
BAR-RAG 2026.02		46.1
RAG SFT 2026.02		45.8
BAR-RAG 2026.02		44.7
O2-Searcher-instruct 2026.04		44.4
RAG w/Reranker 2026.02		43.6
Llama3-Instruct 70B 2024.07		42.7
RAG SFT 2026.02		42.7
RAG 2026.02		42.7
Llama3-ChatQA-1.5 8B 2024.07		42.4
Search-R1-base 2026.04		42.1
InForage-instruct 2026.04		42.1
BAR-RAG 2026.02		42
GPT-4-turbo-2024-0409 2024.07		41.5
BAR-RAG 2026.02		41.5
ZeroSearch-instruct 2026.04		41.4
RAG w/Reranker 2026.02		40.5
GPT-4-0613 RAG 2024.07		40.4
GPT-4-0613 2024.07		40.3
GPT-4-turbo-2024-0409 RAG 2024.07		40.3
Mixtral-8x22B-Instruct 2024.07		40.1
BAR-RAG 2026.02		40.1
Query Decomp-base (Ours) 2026.04		40.1
Search-R1-instruct 2026.04		39.7
ReFeed 2024.07		39.6
ZeroSearch-base 2026.04		39.4
RAG 2026.02		39.3
RAG SFT 2026.02		38.9
GPT-3.5-turbo-1106 2024.07		38.6
GLaM 64B 2024.07		37.5
PaLM2 540B 2024.07		37.1
RECOMP 20B 2024.07		37
RAG w/Reranker 2026.02		35.6
RA-DIT 65B 2024.07		35.2
RAG 2026.02		34.8
RAG 2026.04		34.8
GenRead 2024.07		32.5
R-Search-instruct (r.) 2026.04		31.9
Retrieve-Read 2024.07		31.7
Llama3-Instruct 8B 2024.07		30.9
InstructGPT 2024.07		29.9
RePlug 65B 2024.07		28.8
CoT 2026.02		27.8
Atlas 11B 2024.07		26.7
SFT 2026.04		24.9
Search-o1 2026.04		23.8
IRCoT 2026.02		23.5
R1-base 2026.04		22.6
IRCoT 2026.02		22.4
PaLM2 540B 2024.07		21.2
FLAN-LaMDA 137B 2024.07		20.7
Direct Inference 2026.02		18.4
Direct Inference 2026.02		13.4
IRCoT 2026.02		11.1
Direct Inference 2026.02		10.6
Direct Inference 2026.04		10.6
CoT 2026.02		4.8
CoT 2026.02		2.3
CoT 2026.04		2.3