List your company
ResearchFormerly WizWand

Question Answering on TriviaQA (test) (EM)

92.1EMSearch-R1 (EM)
5.7828.1950.673.01Jun 20, 2025Aug 14, 2025Oct 8, 2025Dec 3, 2025Jan 27, 2026Mar 23, 2026May 18, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Search-R1 (EM)Backbone=Qwen3-4B-Instruct-2507, Supervision Strategy=W. Gold Supervision2026.0492.1
TTRLBackbone=Qwen3-4B-Instruct-2507, Supervision Strategy=W.O. Gold Supervision2026.0490.2
CCSBackbone=Qwen3-4B-Instruct-2507, Supervision Strategy=W.O. Gold Supervision2026.0490.2
CCSBackbone=Qwen3-32B, Supervision Strategy=W.O. Gold Supervision2026.0489.6
CJBackbone=Qwen3-4B-Instruct-2507, Supervision Strategy=W.O. Gold Supervision2026.0489.3
RLIFBackbone=Qwen3-32B, Supervision Strategy=W.O. Gold Supervision2026.0488.7
Search-o1Backbone=Qwen3-32B, Supervision Strategy=Model Inference2026.0488.1
CCSBackbone=Qwen2.5-7B-Instruct, Supervision Strategy=W.O. Gold Supervision2026.0488
Search-R1 (EM)Backbone=Qwen3-32B, Supervision Strategy=W. Gold Supervision2026.0488
RLIFBackbone=Qwen3-4B-Instruct-2507, Supervision Strategy=W.O. Gold Supervision2026.0487.9
Search-R1 (EM)Backbone=Qwen2.5-7B-Instruct, Supervision Strategy=W. Gold Supervision2026.0487.2
TTRLBackbone=Qwen3-32B, Supervision Strategy=W.O. Gold Supervision2026.0487
CJBackbone=Qwen2.5-7B-Instruct, Supervision Strategy=W.O. Gold Supervision2026.0486.5
RLIFBackbone=Qwen2.5-7B-Instruct, Supervision Strategy=W.O. Gold Supervision2026.0484.6
RAGBackbone=Qwen3-32B, Supervision Strategy=Model Inference2026.0483.8
RAGBackbone=Qwen2.5-7B-Instruct, Supervision Strategy=Model Inference2026.0480.5
RAGBackbone=Qwen3-4B-Instruct-2507, Supervision Strategy=Model Inference2026.0480.3
IRCoTBackbone=Qwen3-4B-Instruct-2507, Supervision Strategy=Model Inference2026.0477
IRCoTBackbone=Qwen2.5-7B-Instruct, Supervision Strategy=Model Inference2026.0476.1
ICRLModel=Qwen2.5-14B2026.0375
IRCoTBackbone=Qwen3-32B, Supervision Strategy=Model Inference2026.0469.7
SFTBackbone=Qwen3-32B, Supervision Strategy=W. Gold Supervision2026.0467
INSPOBackbone=Qwen-2.5-7B, Category=Tool-integrated Methods2025.1264.6
MR-Search-BaseTraining Pipeline=Pure RL, External Teacher=–, Retrieval Strategy=w/ Multi-Hop Retrieval, Backbone=Qwen2.5-3B2026.0563.5
SD-Search-BaseTraining Pipeline=Pure RL, External Teacher=–, Retrieval Strategy=w/ Multi-Hop Retrieval, Backbone=Qwen2.5-3B2026.0562.4
SD-Search-InstructTraining Pipeline=Pure RL, External Teacher=–, Retrieval Strategy=w/ Multi-Hop Retrieval, Backbone=Qwen2.5-3B2026.0562.4
SD-Search-BaseBase Model=Qwen2.5-3B, Number of Seeds=52026.0562.4
AutoRefine-BaseBase Model=Qwen2.5-3B, Number of Seeds=52026.0562.1
PersonalAILLM=DeepSeek V3, Retrieval algorithm=hybrid of BeamSearch and WaterCircles, Graph restriction=no restrictions applied2025.0662
AutoRefine-BaseTraining Pipeline=Pure RL, External Teacher=–, Retrieval Strategy=w/ Multi-Hop Retrieval, Backbone=Qwen2.5-3B2026.0562
Search-R1Backbone=Qwen-2.5-3B, Category=Tool-integrated Methods2025.1259.9
Thinker-InstructTraining Pipeline=Synthesize → SFT, External Teacher=Qwen2.5-72B, Retrieval Strategy=w/ Multi-Hop Retrieval, Backbone=Qwen2.5-3B2026.0559.8
ReSearch-BaseTraining Pipeline=Pure RL, External Teacher=–, Retrieval Strategy=w/ Multi-Hop Retrieval, Backbone=Qwen2.5-3B2026.0559.7
AutoRefine-InstructTraining Pipeline=Pure RL, External Teacher=–, Retrieval Strategy=w/ Multi-Hop Retrieval, Backbone=Qwen2.5-3B2026.0559.7
Thinker-InstructBase Model=Qwen2.5-3B, Number of Seeds=52026.0559.4
SFTBackbone=Qwen-2.5-7B, Category=Tool-integrated Methods2025.1259.2
Search-o1Backbone=Qwen2.5-7B-Instruct, Supervision Strategy=Model Inference2026.0458.8
INSPOBackbone=Qwen-2.5-3B, Category=Tool-integrated Methods2025.1258.7
SFTBackbone=Qwen2.5-7B-Instruct, Supervision Strategy=W. Gold Supervision2026.0458.7
RAGBackbone=Qwen-2.5-7B, Category=Tool-integrated Methods2025.1258.5
Search-R1-BaseTraining Pipeline=Pure RL, External Teacher=–, Retrieval Strategy=w/ Multi-Hop Retrieval, Backbone=Qwen2.5-3B2026.0558.3
Search-R1Backbone=Qwen-2.5-7B, Category=Tool-integrated Methods2025.1257.5
ReSearch-InstructTraining Pipeline=Pure RL, External Teacher=–, Retrieval Strategy=w/ Multi-Hop Retrieval, Backbone=Qwen2.5-3B2026.0557.1
Search-R1-InstructTraining Pipeline=Pure RL, External Teacher=–, Retrieval Strategy=w/ Multi-Hop Retrieval, Backbone=Qwen2.5-3B2026.0556.5
CoTModel=Qwen2.5-14B2026.0356.4
CJBackbone=Qwen3-32B, Supervision Strategy=W.O. Gold Supervision2026.0455.6
RAGBackbone=Qwen-2.5-3B, Category=Tool-integrated Methods2025.1254.4
Naive RAGTraining Pipeline=Zero-Shot, External Teacher=–, Retrieval Strategy=w/ Single-Hop Retrieval, Backbone=Qwen2.5-3B2026.0554.4
GRPOBackbone=Qwen-2.5-7B, Category=Model Inference2025.1253.9
DirectModel=Qwen2.5-14B2026.0352
TTRLBackbone=Qwen2.5-7B-Instruct, Supervision Strategy=W.O. Gold Supervision2026.0450
Search-o1Backbone=Qwen3-4B-Instruct-2507, Supervision Strategy=Model Inference2026.0448.9
SFTBackbone=Qwen-2.5-3B, Category=Tool-integrated Methods2025.1248.8
IRCoTBackbone=Qwen-2.5-7B, Category=Tool-integrated Methods2025.1247.8
SFTBackbone=Qwen3-4B-Instruct-2507, Supervision Strategy=W. Gold Supervision2026.0447.4
Search-o1Backbone=Qwen-2.5-3B, Category=Tool-integrated Methods2025.1247.2
Search-o1Training Pipeline=Prompting, External Teacher=–, Retrieval Strategy=w/ Multi-Hop Retrieval, Backbone=Qwen2.5-3B2026.0547.2
PersonalAILLM=Llama3.1 8B, Retrieval algorithm=BeamSearch, Graph restriction=episodic vertices excluded2025.0647
GRPOBackbone=Qwen-2.5-3B, Category=Model Inference2025.1245.5
R1-BaseTraining Pipeline=Pure RL, External Teacher=–, Retrieval Strategy=w/o Retrieval, Backbone=Qwen2.5-3B2026.0545.5
T2MBackbone=LLaDA2.1-mini (16B MoE), Decoding Strategy=greedy decoding, Temperature=0, block length=32, LOWPROB parameters=τ=0.3, Cmax=1, ρmax=0.252026.0444.98
R1-InstructTraining Pipeline=Pure RL, External Teacher=–, Retrieval Strategy=w/o Retrieval, Backbone=Qwen2.5-3B2026.0544.9
Search-o1Backbone=Qwen-2.5-7B, Category=Tool-integrated Methods2025.1244.3
T2TBackbone=LLaDA2.1-mini (16B MoE), Decoding Strategy=greedy decoding, Temperature=0, block length=322026.0443.71
DirectBackbone=Qwen-2.5-7B, Category=Model Inference2025.1240.8
SFTBackbone=Qwen-2.5-7B, Category=Model Inference2025.1235.4
CoTBackbone=Qwen3-32B, Supervision Strategy=Model Inference2026.0432.2
IRCoTBackbone=Qwen-2.5-3B, Category=Tool-integrated Methods2025.1231.2
IRCoTTraining Pipeline=Prompting, External Teacher=–, Retrieval Strategy=w/ Multi-Hop Retrieval, Backbone=Qwen2.5-3B2026.0531.2
Direct InferenceBackbone=Qwen3-32B, Supervision Strategy=Model Inference2026.0430.8
SFTBackbone=Qwen-2.5-3B, Category=Model Inference2025.1229.2
SFTTraining Pipeline=SFT, External Teacher=–, Retrieval Strategy=w/o Retrieval, Backbone=Qwen2.5-3B2026.0529.2
DirectBackbone=Qwen-2.5-3B, Category=Model Inference2025.1228.8
Direct GenerationTraining Pipeline=Zero-Shot, External Teacher=–, Retrieval Strategy=w/o Retrieval, Backbone=Qwen2.5-3B2026.0528.8
PersonalAILLM=Qwen2.5 7B, Retrieval algorithm=BeamSearch, Graph restriction=episodic vertices excluded2025.0620
PersonalAILLM=DeepSeek R1 7B, Retrieval algorithm=A*, Graph restriction=episodic vertices excluded2025.0618
CoTBackbone=Qwen3-4B-Instruct-2507, Supervision Strategy=Model Inference2026.0414.2
CoTBackbone=Qwen2.5-7B-Instruct, Supervision Strategy=Model Inference2026.0413.9
Direct InferenceBackbone=Qwen2.5-7B-Instruct, Supervision Strategy=Model Inference2026.0411.1
Direct InferenceBackbone=Qwen3-4B-Instruct-2507, Supervision Strategy=Model Inference2026.049.1