List your company

Question Answering on NQ, TriviaQA, PopQA (test)

51.3Overall Average ScoreDART
-0.49212.95426.439.846Dec 3, 2025Dec 13, 2025Dec 23, 2025Jan 2, 2026Jan 12, 2026Jan 22, 2026Feb 1, 2026

Evaluation results

MethodLinks
DARTBackbone=Qwen2.5-3b-Base2026.0251.345.760.547.8—
DARTBackbone=Qwen2.5-3b-Instruct2026.025145.160.247.6—
Search-R1-GRPOBackbone=Qwen2.5-3b-Base2026.0247.84458.241.3—
Search-R1-PPO-BaseBackbone=Qwen2.5-3b-Base2026.0247.640.658.743.5—
Search-R1-GRPOBackbone=Qwen2.5-3b-Instruct2026.0245.139.756.539.1—
LLDSBackbone=Qwen2.5-3b-Instruct, Base Training=Search-R1-GSPO2025.1245.148.96448.353.7
LLDSBackbone=Qwen2.5-3b-Instruct, Base Training=Search-R1-GRPO2025.1244.146.262.746.851.9
LLDS-MABackbone=Qwen2.5-3b-Base, Base Training=Search-R1-GRPO2025.124448.363.347.453
RAGBackbone=Qwen2.5-3B2026.0242.634.854.438.7—
Tree-GRPOBackbone=Qwen2.5-3b-Base/Instruct2025.1242.446.859.743.650
Search-R1-PPO-InsBackbone=Qwen2.5-3b-Instruct2026.0242.134.154.537.8—
DARTBackbone=Qwen2.5-3b-Base2026.0240.5————
DARTBackbone=Qwen2.5-3b-Instruct2026.0239.9————
Rejection SamplingBackbone=Qwen2.5-3B2026.0237.129.448.833.2—
LLDSBackbone=Qwen2.5-3b-Base, Base Training=Search-R1-GRPO2025.123647.962.745.352
ZeroSearch-BaseBackbone=Qwen2.5-3b-Base/Instruct2025.1234.54361.641.448.7
Search-R1-GRPOBackbone=Qwen2.5-3b-Instruct2026.0233.6————
Search-R1-GRPOBackbone=Qwen2.5-3b-Instruct2025.1233.639.756.539.145.1
Search-R1-GSPOBackbone=Qwen2.5-3b-Instruct2025.1233.137.656.137.443.7
Search-R1-PPO-InsBackbone=Qwen2.5-3b-Instruct2026.0232.5————
Search-R1-PPO-InsBackbone=Qwen2.5-3b-Base/Instruct2025.1232.534.154.537.842.1
Search-o1Backbone=Qwen2.5-3B2026.0232.423.847.226.2—
ZeroSearch-InsBackbone=Qwen2.5-3b-Base/Instruct2025.1231.741.457.444.847.9
Search-R1-GRPOBackbone=Qwen2.5-3b-Base2025.1231.242.158.341.347.2
StepSearchBackbone=Qwen2.5-3b-Base/Instruct2025.1231.129.64934.137.5
Search-R1-PPO-BaseBackbone=Qwen2.5-3b-Base2026.0230.3————
Search-R1-GRPOBackbone=Qwen2.5-3b-Base2026.0230.3————
Search-R1-PPO-BaseBackbone=Qwen2.5-3b-Base/Instruct2025.1230.340.658.743.547.6
R1-baseBackbone=Qwen2.5-3b-Base2026.0228.522.645.517.3—
R1-instructBackbone=Qwen2.5-3b-Instruct2026.0227.72144.917.1—
RAGBackbone=Qwen2.5-3B2026.0227————
RAGBackbone=Qwen2.5-3b-Base/Instruct2025.122734.854.438.742.6
Rejection SamplingBackbone=Qwen2.5-3B2026.0226.5————
Rejection SamplingBackbone=Qwen2.5-3b-Base/Instruct2025.1226.529.448.833.237.1
Search-o1Backbone=Qwen2.5-3B2026.0225.5————
Search-o1Backbone=Qwen2.5-3b-Base/Instruct2025.1225.523.847.226.232.4
R1-baseBackbone=Qwen2.5-3b-Base2026.0222.9————
R1-baseBackbone=Qwen2.5-3b-Base/Instruct2025.1222.922.645.517.328.5
R1-instructBackbone=Qwen2.5-3b-Instruct2026.0222.4————
R1-instructBackbone=Qwen2.5-3b-Base/Instruct2025.1222.42144.917.127.7
SFTBackbone=Qwen2.5-3B2026.0221.524.929.210.4—
IRCoTBackbone=Qwen2.5-3B2026.0220.811.131.220—
IRCoTBackbone=Qwen2.5-3B2026.0218.1————
SFTBackbone=Qwen2.5-3B2026.0217.6————
Direct InferenceBackbone=Qwen2.5-3B2026.0216.710.628.810.8—
Direct InferenceBackbone=Qwen2.5-3B2026.0213.4————
Direct InferenceBackbone=Qwen2.5-3b-Base/Instruct2025.1213.410.628.810.816.7
CoTBackbone=Qwen2.5-3B2026.0222.33.20.5—
CoTBackbone=Qwen2.5-3B2026.021.5————