Share your thoughts, 1 month free Claude Pro on usSee more

Mathematical Reasoning on GaoKao En 2023

79.9Pass@1 Accuracy

CES

Updated 8d ago

Evaluation Results

Method
CES 2026.05	79.9	1,914
Pass@8 (Upper Bound) 2025.05	79.7	-
GPT-o1-mini 2025.02	78.4	-
Scaf-GRPO 2025.10	72.3	-
SCOPE (w/o AST normalization) 2025.05	72.2	-
DAPO 2026.05	72.2	2,182
SCOPE 2025.05	71.9	-
R1-7B 2026.05	71.8	2,511
Entropy Shape 2026.05	71.8	2,048
SCOPE (Step Skipping) 2025.05	71.7	-
Vanilla GRPO 2025.10	71.4	-
SCOPE (Step Replacement) 2025.05	71	-
RLHFlow-PRM-Mistral-8B 2025.05	70.9	-
Skywork-PRM-7B 2025.05	70.9	-
Skywork-PRM-1.5B 2025.05	70.6	-
SCOPE (w/o code translation) 2025.05	70.6	-
Qwen2.5-Math-7B-PRM800K 2025.05	70.5	-
Majority @8 2025.05	70.4	-
Qwen2.5-Math-7B-S2R-ORL 2025.02	70.1	-
RLHFlow-PRM-Deepseek-8B 2025.05	69.9	-
OpenReasoner-Zero-7B 2025.06	68.8	-
GPT-4o 2025.02	67.5	-
EurusPRM-Stage1 2025.05	66.1	-
EurusPRM-Stage2 2025.05	65.8	-
Uni-DPO 2025.06	65.7	-
Uni-DPO 2025.06	65.7	-
SimpleRL-Zero-7B 2025.06	65.5	-
Dr. GRPO (Oat-Zero-7B) 2025.06	64.9	-
Greedy 2025.05	64.2	-
Vanilla GRPO 2025.10	64.2	-
Scaf-GRPO 2025.10	63.8	-
Math-Shepherd-PRM-7B 2025.05	63.6	-
SFT 2025.06	63.6	-
RAFT 2025.06	63.6	-
Scaf-GRPO 2025.10	63.4	-
Iterative DPO 2025.06	63.4	-
Oat-Zero 2025.10	62.9	-
LUFFY 2025.10	62.7	-
Vanilla GRPO 2025.10	62.6	-
DeepSeek-R1-Distill-Qwen-1.5B 2025.10	62.1	-
SimPO 2025.06	62.1	-
PPO 2025.06	62.1	-
SimPO 2025.06	62.1	-
DPO 2025.06	61.6	-
DPO 2025.06	61.6	-
Eurus2-PRIME 2025.06	61.3	-
SimpleRL-Zero 2025.10	60.8	-
CES 2026.05	60.8	3,173
Uni-DPO 2025.06	58.2	-
Scaf-GRPO 2025.10	57.9	-
DAPO 2026.05	57.7	3,304
Vanilla GRPO 2025.10	57.4	-
DPO 2025.06	56.6	-
SimPO 2025.06	55.3	-
Baseline 2025.06	53	-
Eurus2-SFT 2025.06	52.2	-
Qwen2-7B-S2R-ORL 2025.02	50.9	-
Scaf-GRPO 2025.10	46	-
Vanilla GRPO 2025.10	45.7	-
Llama-3.1-8B-S2R-ORL 2025.02	45.2	-
Baseline 2025.06	44.7	-
Baseline 2025.06	44.7	-
Qwen2.5-7B 2025.10	42.6	-
Qwen2.5-Math-7B 2025.10	35.1	-
Llama-3.2-3B-Instruct 2025.10	33.5	-
Qwen2.5-Math-1.5B 2025.10	20	-