Agentic Search on DeepSearchQA
90AccuracyFINEVERIFY
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FINEVERIFYModel=Gemini-3-flash, Sampling Trajectories=42026.05 | 90 | — | |
| Best-of-NModel=Gemini-3-flash, Sampling Trajectories=42026.05 | 89.5 | — | |
| Confidence VerifyModel=Gemini-3-flash, Sampling Trajectories=42026.05 | 89.5 | — | |
| Solution AggregationModel=Gemini-3-flash, Sampling Trajectories=42026.05 | 89 | — | |
| Weighted VotingModel=Gemini-3-flash, Sampling Trajectories=42026.05 | 88.5 | — | |
| Majority VotingModel=Gemini-3-flash, Sampling Trajectories=42026.05 | 85.5 | — | |
| Pass@1Model=Gemini-3-flash, Sampling Trajectories=12026.05 | 84 | — | |
| Weighted VotingModel=GPT-5-mini, Sampling Trajectories=42026.05 | 77 | — | |
| FINEVERIFYModel=GPT-5-mini, Sampling Trajectories=42026.05 | 77 | — | |
| Best-of-NModel=GPT-5-mini, Sampling Trajectories=42026.05 | 76 | — | |
| Solution AggregationModel=GPT-5-mini, Sampling Trajectories=42026.05 | 76 | — | |
| Confidence VerifyModel=GPT-5-mini, Sampling Trajectories=42026.05 | 74.5 | — | |
| Majority VotingModel=GPT-5-mini, Sampling Trajectories=42026.05 | 72 | — | |
| Pass@1Model=GPT-5-mini, Sampling Trajectories=12026.05 | 70.5 | — | |
| SELFCOMPACTBackbone=MiniMax-M2.52026.06 | 61.3 | 0.08 | |
| Keep-last-NBackbone=MiniMax-M2.52026.06 | 57 | 0.06 | |
| SELFCOMPACTBackbone=Mimo-V2-Flash2026.06 | 56.8 | 0.13 | |
| Fixed-intervalBackbone=MiniMax-M2.52026.06 | 56.7 | 0.06 | |
| Delete-allBackbone=MiniMax-M2.52026.06 | 54.9 | 0.04 | |
| Keep-last-NBackbone=Mimo-V2-Flash2026.06 | 53 | 0.1 | |
| Fixed-intervalBackbone=Mimo-V2-Flash2026.06 | 52.3 | 0.14 | |
| No CompactionBackbone=MiniMax-M2.52026.06 | 52 | 0.19 | |
| Delete-allBackbone=Mimo-V2-Flash2026.06 | 49.7 | 0.09 | |
| No CompactionBackbone=Mimo-V2-Flash2026.06 | 46.5 | 0.25 | |
| SELFCOMPACTBackbone=GLM-4.7-Flash2026.06 | 44 | 0.07 | |
| Fixed-intervalBackbone=GLM-4.7-Flash2026.06 | 39.1 | 0.05 | |
| Keep-last-NBackbone=GLM-4.7-Flash2026.06 | 38.8 | 0.04 | |
| Delete-allBackbone=GLM-4.7-Flash2026.06 | 35.8 | 0.03 | |
| No CompactionBackbone=GLM-4.7-Flash2026.06 | 34.2 | 0.13 |