List your company
ResearchFormerly WizWand

Question Answering on PopQA (Exact Match)

64.2Exact MatchMASS-RAG-Llama3 8B
23.057633.738844.4255.1012Aug 27, 2025Oct 10, 2025Nov 24, 2025Jan 7, 2026Feb 21, 2026Apr 6, 2026May 21, 2026
Updated 5mo ago

Evaluation results

MethodLinks
MASS-RAG-Llama3 8BRetrieval Strategy=Proposed2026.0464.2
MAIN-RAG-Llama3 8BRetrieval Strategy=Training-free retrieval2026.0464
MASS-RAG-Qwen3 8BRetrieval Strategy=Proposed2026.0463.9
Qwen3 8BRetrieval Strategy=Training-free retrieval2026.0462.1
Llama3 8BRetrieval Strategy=Training-free retrieval2026.0461.8
MASS-RAG-Mistral 7BRetrieval Strategy=Proposed2026.0460.9
MAIN-RAG-Mistral 7BRetrieval Strategy=Training-free retrieval2026.0458.9
MASS-RAG-Llama2 7BRetrieval Strategy=Proposed2026.0457.8
Mistral 7BRetrieval Strategy=Training-free retrieval2026.0455.5
Self-RAG 7BRetrieval Strategy=Training-based retrieval2026.0454.9
ZeroSearchBackbone=Qwen2.5-7B Instruct2026.0551.5
Llama2 7BRetrieval Strategy=Training-free retrieval2026.0450.9
AMATA 8BModel Scale=8B, Backbone Architecture=AMATA2026.0549.62
ZeroSearchParameter Scale=7B, Search Environment=Google Web Search2025.1148.8
MemSearcherParameter Scale=14B2025.1148.8
Llama2-FT 7BRetrieval Strategy=Training-based retrieval2026.0448.7
ReSearchParameter Scale=32B2025.1148.2
Search-R1-baseParameter Scale=14B2025.1148
MemSearcherParameter Scale=3B2025.1147.9
MemSearcherParameter Scale=7B2025.1147.8
SMART 8BModel Scale=8B, Backbone Architecture=SMART2026.0547.66
EvolveRBackbone Model=Qwen2.5-7B2025.1047.3
GiGPO 8BModel Scale=8B, Backbone Architecture=GiGPO2026.0547.29
SLATEBackbone=Qwen2.5-7B-Base2026.0247
SPA-RL 8BModel Scale=8B, Backbone Architecture=SPA-RL2026.0546.85
Search-E1-InstructRetrieval Strategy=Multi-Hop Retrieval, Supervision Paradigm=Process-supervision RL2026.0546.1
SEARCH-R1Backbone=Qwen2.5-7B-Base2026.0245.7
Llama2 13BRetrieval Strategy=Training-free retrieval2026.0445.7
Search-R1-baseParameter Scale=7B2025.1145.7
SLATEBackbone=Qwen2.5-3B-Base, Training Data=NQ+HotpotQA2026.0245.1
AutoRefine-baseParameter Scale=3B2025.1145
TIPSBackbone=Qwen2.5-7B Instruct2026.0545
AutoRefine-BaseRetrieval Strategy=Multi-Hop Retrieval, Supervision Paradigm=Outcome-reward RL2026.0545
ZeroSearchParameter Scale=3B, Search Environment=Google Web Search2025.1144.8
ZeroSearchBackbone=Qwen2.5-3B Instruct2026.0544.8
AutoRefine-instructParameter Scale=3B2025.1144.7
AutoRefine-InstructRetrieval Strategy=Multi-Hop Retrieval, Supervision Paradigm=Outcome-reward RL2026.0544.7
ReSearchParameter Scale=7B2025.1144.6
Search-R1-instructParameter Scale=14B2025.1144.2
PiCABackbone=Qwen2.5-7B Instruct2026.0544.2
SelfRag 8BModel Scale=8B, Backbone Architecture=Self-RAG2026.0543.55
SEARCH-R1Backbone=Qwen2.5-3B-Base, Training Data=NQ+HotpotQA2026.0243.5
Search-R1-baseParameter Scale=3B2025.1143.5
Search-R1-baseBackbone Model=Qwen2.5-3B2025.1043.5
EvolveRBackbone Model=Qwen2.5-3B2025.1043.4
ReSearch-BaseRetrieval Strategy=Multi-Hop Retrieval, Supervision Paradigm=Outcome-reward RL2026.0543
O2-SearcherParameter Scale=3B2025.1142.9
TIPSBackbone=Qwen2.5-3B Instruct2026.0542.8
GiGPO-InstructRetrieval Strategy=Multi-Hop Retrieval, Supervision Paradigm=Process-supervision RL2026.0542.4
PiCABackbone=Qwen2.5-3B Instruct2026.0541.7
MT-PPOBackbone=Qwen2.5-7B Instruct2026.0541.6
Search-R1-BaseRetrieval Strategy=Multi-Hop Retrieval, Supervision Paradigm=Outcome-reward RL2026.0541.3
RADIT 8BModel Scale=8B, Backbone Architecture=RADIT2026.0541.15
R1-SearcherParameter Scale=7B, Search Environment=Google Web Search2025.1141
MT-PPOBackbone=Qwen2.5-3B Instruct2026.0540.5
DGPOModel Family=Qwen 2.5, Teacher Model=Qwen 2.5 3B, Student Model=Qwen 2.5 0.5B2025.0840.2
Search-R1-instructParameter Scale=7B2025.1139.7
Search-R1-instructBackbone Model=Qwen2.5-7B2025.1039.7
Search-R1Backbone=Qwen2.5-7B Instruct2026.0539.7
ReSearch-InstructRetrieval Strategy=Multi-Hop Retrieval, Supervision Paradigm=Outcome-reward RL2026.0539.5
Teacher-3BModel Family=Qwen 2.5, Teacher Model=Qwen 2.5 3B2025.0839.3
RAGBackbone Model=Qwen2.5-7B2025.1039.2
RAGBackbone=Qwen2.5-7B Instruct2026.0539.2
Search-R1-InstructRetrieval Strategy=Multi-Hop Retrieval, Supervision Paradigm=Outcome-reward RL2026.0539.1
ZeroSearchBackbone=Qwen2.5-7B-Base2026.0238.9
RAGBackbone Model=Qwen2.5-3B2025.1038.7
RAGBackbone=Qwen2.5-3B Instruct2026.0538.7
Naive RAGRetrieval Strategy=Single-Hop Retrieval2026.0538.7
Rejection SamplingBackbone Model=Qwen2.5-7B2025.1038
PPOModel Family=Qwen 2.5, Teacher Model=Qwen 2.5 3B, Student Model=Qwen 2.5 0.5B2025.0837.9
Search-R1-instructParameter Scale=3B2025.1137.8
Search-R1-instructBackbone Model=Qwen2.5-3B2025.1037.8
Search-R1Backbone=Qwen2.5-3B Instruct2026.0537.8
KDModel Family=Qwen 2.5, Teacher Model=Qwen 2.5 3B, Student Model=Qwen 2.5 0.5B2025.0837.3
DistiLLMModel Family=Qwen 2.5, Teacher Model=Qwen 2.5 3B, Student Model=Qwen 2.5 0.5B2025.0837.3
ReSearchBackbone=Qwen2.5-7B-Base2026.0237.2
ZeroSearchBackbone=Qwen2.5-3B-Base2026.0236.5
TAIDModel Family=Qwen 2.5, Teacher Model=Qwen 2.5 3B, Student Model=Qwen 2.5 0.5B2025.0836.5
SeqKDModel Family=Qwen 2.5, Teacher Model=Qwen 2.5 3B, Student Model=Qwen 2.5 0.5B2025.0836.4
SFT → PPOModel Family=Qwen 2.5, Teacher Model=Qwen 2.5 3B, Student Model=Qwen 2.5 0.5B2025.0835.9
GKDModel Family=Qwen 2.5, Teacher Model=Qwen 2.5 3B, Student Model=Qwen 2.5 0.5B2025.0835.8
ReSearchBackbone=Qwen2.5-3B-Base2026.0234.2
RAGBackbone=Qwen2.5-7B-Base2026.0233.9
Rejection SamplingBackbone Model=Qwen2.5-3B2025.1033.2
SFTBackbone=Qwen2.5-7B-Base2026.0232.5
IRCoTBackbone=Qwen2.5-7B-Base2026.0231.2
GPT4oBackbone Architecture=GPT-4o2026.0530.94
IRCoTBackbone Model=Qwen2.5-7B2025.1030.1
Search-o1Backbone=Qwen2.5-7B-Base2026.0229.8
RAGBackbone=Qwen2.5-3B-Base2026.0229.6
Llama3 8BRetrieval Strategy=Without retrieval2026.0429.2
SFTBackbone=Qwen2.5-3B-Base2026.0228.6
R1 (no search)Backbone=Qwen2.5-7B-Base2026.0227.8
IRCoTBackbone=Qwen2.5-3B-Base2026.0226.8
Mistral 7BRetrieval Strategy=Without retrieval2026.0426.2
Search-o1Backbone Model=Qwen2.5-3B2025.1026.2
Search-o1Backbone=Qwen2.5-3B Instruct2026.0526.2
Search-o1Retrieval Strategy=Multi-Hop Retrieval, Supervision Paradigm=Prompting2026.0526.2
Search-o1Backbone=Qwen2.5-3B-Base2026.0225.4
Qwen2.5-VL 3BText Source=Pure-Text, CEPE protocol=true2025.1024.64