List your company
ResearchFormerly WizWand

Multi-Hop Question Answering on 2Wiki (out-of-domain)

47.6AccuracyReSearch-7B-Instruct
9.6419.49529.3539.205Oct 15, 2025Nov 3, 2025Nov 23, 2025Dec 12, 2025Jan 1, 2026Jan 20, 2026Feb 9, 2026
Updated 5mo ago

Evaluation results

MethodLinks
ReSearch-7B-InstructModel Scale=7B, Model Variant=Instruct2025.1047.6
ReSearch-7B-BaseModel Scale=7B, Model Variant=Base2025.1044.7
EM+Info-ThinkFramework=VERITAS-R1, Reward Components=EM+Info-Think2025.1042.3
EvolveRBackbone=Qwen2.5-7B-Instruct2026.0242
EM+Think-AnsFramework=VERITAS-R1, Reward Components=EM+Think-Ans2025.1042
EM+Info-Think+Think-AnsFramework=VERITAS-R1, Reward Components=EM+Info-Think+Think-Ans2025.1041.9
Search-R1-7B-Base-PPO w/ FormatModel Scale=7B, Training Algorithm=PPO, Input Formatting=Included2025.1041.2
SKILLRLBackbone=Qwen2.5-7B-Instruct2026.0240.3
Search-R1Backbone=Qwen2.5-7B-Instruct2026.0240.1
Search-R1-7B-Base-PPOModel Scale=7B, Training Algorithm=PPO2025.1038.2
DeSA-7B-Instrct-GRPOModel Scale=7B, Training Algorithm=GRPO2025.1037.4
StepSearchBackbone=Qwen2.5-7B-Instruct2026.0236.6
DeSA-3B-Instruct-GRPOModel Scale=3B, Training Algorithm=GRPO2025.1036.3
ZeroSearchBackbone=Qwen2.5-7B-Instruct2026.0235.2
Rejection SamplingReasoning Strategy=Rejection Sampling2025.1029.6
R1-instructModel Variant=Instruct2025.1029.2
R1-InstructBackbone=Qwen2.5-7B-Instruct2026.0227.5
R1-baseModel Variant=Base2025.1027.3
Search-o1Backbone=Qwen2.5-7B-Instruct2026.0227
SFTTraining Algorithm=Supervised Fine-Tuning2025.1025.9
Direct InferenceReasoning Strategy=Direct Inference2025.1025
RAGReasoning Strategy=Retrieval-Augmented Generation2025.1023.5
RAGBackbone=Qwen2.5-7B-Instruct2026.0223.2
CoTBackbone=Qwen2.5-7B-Instruct2026.0222.6
Qwen2.5Backbone=Qwen2.5-7B-Instruct2026.0222.2
Search-o1Reasoning Strategy=Search-o12025.1017.6
IRCoTReasoning Strategy=Iterative Retrieval Chain-of-Thought2025.1014.9
CoTReasoning Strategy=Chain-of-Thought2025.1011.1