List your company
ResearchFormerly WizWand

Algebraic Reasoning on AQUA

91.89AccuracySafeSieve
23.322841.123958.92576.7261Jul 4, 2024Oct 18, 2024Feb 1, 2025May 19, 2025Sep 2, 2025Dec 17, 2025Apr 3, 2026
Updated 6mo ago

Evaluation results

MethodLinks
SafeSieveParadigm=post-prune, Base model=deepseek-V3-671B2025.0891.89
AgentDropoutParadigm=post-prune, Base model=deepseek-V3-671B2025.0891.37
AgentPruneParadigm=post-prune, Base model=deepseek-V3-671B2025.0890.3
Weighted MV2026.0490.23
Simple MV2026.0490.16
EMS-SimSorting Strategy=Semantic Similarity2026.0490.16
EMS-RelSorting Strategy=Historical Reliability2026.0490.16
Conf-MV2026.0489.92
G-DesignerParadigm=pre-design, Base model=deepseek-V3-671B2025.0889.63
Clean ModelTarget Model=Mistral-7B, Trigger Condition=Without Trigger2025.0488.7
HCP-MADBase Model=Qwen2.5-32b-instruct + Llama-3.1-70b-instruct2026.0488.58
Self-Consistency2026.0488.52
GPTSwarmParadigm=pre-design, Base model=deepseek-V3-671B2025.0888.4
ShadowCoTTarget Model=Mistral-7B, Trigger Condition=Without Trigger2025.0488.3
Single Agent2026.0487.66
DarkMindTarget Model=Mistral-7B, Trigger Condition=Without Trigger2025.0487.4
BadChainTarget Model=Mistral-7B, Trigger Condition=Without Trigger2025.0487
MADBase Model=Llama-3.1-70b-instruct2026.0486.61
MADBase Model=Qwen2.5-32b-instruct2026.0485.83
SCBase Model=Llama-3.1-70b-instruct2026.0485.43
CoTParadigm=single, Base model=deepseek-V3-671B2025.0885.42
SCBase Model=Qwen2.5-32b-instruct2026.0485.04
SafeSieveParadigm=post-prune, Base model=gpt-4o-mini (~8B)2025.0884.9
AgentDropoutParadigm=post-prune, Base model=gpt-4o-mini (~8B)2025.0884.87
AgentPruneParadigm=post-prune, Base model=gpt-4o-mini (~8B)2025.0884.71
CoTBase Model=Llama-3.1-70b-instruct2026.0484.65
DOWNBase Model=Qwen2.5-32b-instruct + Llama-3.1-70b-instruct2026.0484.65
VanillaParadigm=single, Base model=deepseek-V3-671B2025.0884.58
Graph-GRPO2026.0384.21
Heter-MADBase Model=Qwen2.5-32b-instruct + Llama-3.1-70b-instruct2026.0483.86
EIB-LEARNER2026.0383.49
MARSBase Model=Qwen2.5-32b-instruct + Llama-3.1-70b-instruct2026.0483.07
SRBase Model=Qwen2.5-32b-instruct2026.0482.68
SRBase Model=Llama-3.1-70b-instruct2026.0482.28
G-designer2026.0381.6
AgentDropout2026.0380.94
AgentPrune2026.0380.51
CoTBase Model=Qwen2.5-32b-instruct2026.0480.31
G-DesignerParadigm=pre-design, Base model=gpt-4o-mini (~8B)2025.0880.07
QAP50Model=GPT-4 Turbo, Constraint (N)=502024.0779.1
BaselineModel=GPT-4 Turbo2024.0778.7
TADBModel=GPT-4 Turbo2024.0778.7
GPTSwarmParadigm=pre-design, Base model=gpt-4o-mini (~8B)2025.0878.4
QAP150Model=GPT-4 Turbo, Constraint (N)=1502024.0778
LLM-Debate2026.0377.65
QAP25Model=GPT-4 Turbo, Constraint (N)=252024.0777.6
Random2026.0376.48
QAP200Model=GPT-4 Turbo, Constraint (N)=2002024.0776.4
SC (CoT)2026.0375.63
QAP100Model=GPT-4 Turbo, Constraint (N)=1002024.0775.6
CoTModel=GPT-4 Turbo2024.0774.4
Chain2026.0374.05
CoT2026.0373.58
Complete2026.0372.95
VanillaParadigm=single, Base model=gpt-4o-mini (~8B)2025.0871.42
Tree2026.0371.23
Vanilla2026.0371.06
CoTParadigm=single, Base model=gpt-4o-mini (~8B)2025.0865
PS+Model=GPT-4 Turbo2024.0752.8
Qwen3-8BCategory=Naive LLM2026.0133.45
VIST2-8BCategory=Ours2026.0132.1
Qwen3-VL-8BCategory=Visual-enhanced LLM2026.0129.88
Qwen3-4BCategory=Naive LLM2026.0128.4
VIST2-4BCategory=Ours2026.0127.95
Qwen3-VL-4BCategory=Visual-enhanced LLM2026.0125.96