List your company
ResearchFormerly WizWand

Math Problem Solving on AIME 24

90AccuracyBERT-Judge
-2.736821.339145.41569.4909Feb 3, 2026Feb 26, 2026Mar 21, 2026Apr 13, 2026May 6, 2026May 29, 2026Jun 21, 2026
Updated 4mo ago

Evaluation results

MethodLinks
BERT-Judgeclean_name=BERT-as-a-Judge2026.0490
Regex2026.0487.8
LLM-Judgebackbone=Qwen-3 0.6B2026.0477.9
Primitives-based MASModel=Qwen3-8B2026.0276.7
Primitives-based MASModel=DeepSeek-R1-Distill Llama-70B2026.0276.7
SingleModel=DeepSeek-R1-Distill Qwen-32B2026.0273.3
TextMASModel=DeepSeek-R1-Distill Qwen-32B2026.0273.3
LatentMASModel=DeepSeek-R1-Distill Qwen-32B2026.0273.3
Primitives-based MASModel=DeepSeek-R1-Distill Qwen-32B2026.0273.3
VotingModel=DeepSeek-R1-Distill Llama-70B2026.0273.3
VotingModel=Qwen3-8B2026.0270
ReviewModel=DeepSeek-R1-Distill Qwen-32B2026.0270
VotingModel=DeepSeek-R1-Distill Qwen-32B2026.0270
SingleModel=DeepSeek-R1-Distill Llama-70B2026.0270
TextMASModel=DeepSeek-R1-Distill Llama-70B2026.0270
ReviewModel=DeepSeek-R1-Distill Llama-70B2026.0270
PlanningModel=DeepSeek-R1-Distill Qwen-32B2026.0266.7
ReviewModel=Qwen3-8B2026.0263.3
PlanningModel=Qwen3-8B2026.0263.3
PlanningModel=DeepSeek-R1-Distill Llama-70B2026.0263.3
LatentMASModel=Qwen3-8B2026.0256.7
TextMASModel=Qwen3-8B2026.0253.3
SingleModel=Qwen3-8B2026.0250
LatentMASModel=DeepSeek-R1-Distill Llama-70B2026.0243.3
DAPORegime=N-OnP., Backbone=Qwen2.5-7B, Training Dataset=ORZ-57K, Sampling Strategy=avg@322026.0619.69
CISPORegime=N-OnP., Backbone=Qwen2.5-7B, Training Dataset=ORZ-57K, Sampling Strategy=avg@322026.0618.65
DAPORegime=OffP., Backbone=Qwen2.5-7B, Training Dataset=ORZ-57K, Sampling Strategy=avg@322026.0618.02
ACPORegime=N-OnP., Backbone=Qwen2.5-7B, Training Dataset=ORZ-57K, Sampling Strategy=avg@322026.0616.77
ACPORegime=OffP., Backbone=Qwen2.5-7B, Training Dataset=ORZ-57K, Sampling Strategy=avg@322026.0616.35
Low-EntropyRegime=N-OnP., Backbone=Qwen2.5-7B, Training Dataset=ORZ-57K, Sampling Strategy=avg@322026.0616.04
AR-LoptiRegime=N-OnP., Backbone=Qwen2.5-7B, Training Dataset=ORZ-57K, Sampling Strategy=avg@322026.0615.21
High-EntropyRegime=N-OnP., Backbone=Qwen2.5-7B, Training Dataset=ORZ-57K, Sampling Strategy=avg@322026.0614.37
High-EntropyRegime=OffP., Backbone=Qwen2.5-7B, Training Dataset=ORZ-57K, Sampling Strategy=avg@322026.0614.06
CISPORegime=OffP., Backbone=Qwen2.5-7B, Training Dataset=ORZ-57K, Sampling Strategy=avg@322026.0613.44
Low-EntropyRegime=OffP., Backbone=Qwen2.5-7B, Training Dataset=ORZ-57K, Sampling Strategy=avg@322026.0611.15
AR-LoptiRegime=OffP., Backbone=Qwen2.5-7B, Training Dataset=ORZ-57K, Sampling Strategy=avg@322026.068.75
Base ModelRegime=N-OnP., Backbone=Qwen2.5-7B, Training Dataset=ORZ-57K, Sampling Strategy=avg@322026.060.83