Mathematical Reasoning on Math500 (Accuracy, Acc Drop)

94.81Accuracy

FACT-E(Standard)

Updated 4d ago

Evaluation Results

Method
FACT-E(Standard) 2026.04	94.81	-
POLISH 2026.04	94.8	-
FACT-E(Lightweight) 2026.04	94.26	-
CONSISTENCY 2026.04	93.17	-
CoT 2026.04	93.17	-
POLISH 2026.04	92.9	-
REFLECT 2026.04	92.9	-
CONSISTENCY 2026.04	92.9	-
FACT-E(Standard) 2026.04	92.9	-
DENOISE 2026.04	92.08	-
FACT-E(Lightweight) 2026.04	90.32	-
MCTS (Vanilla) 2026.04	88.3	-
MCTS + PE 2026.04	88.3	-
Adaptive Parallel Monte Carlo Tree Search (Full Suite) 2026.04	87.6	-
REFLECT 2026.04	87.43	-
Beam Search 2026.04	87.1	-
FACT-E(Standard) 2026.04	86.61	-
FACT-E(Lightweight) 2026.04	85.52	-
CoT 2026.04	85.52	-
DENOISE 2026.04	84.7	-
POLISH 2026.04	83.33	-
DENOISE 2026.04	83.06	-
CONSISTENCY 2026.04	82.79	-
REFLECT 2026.04	80.6	-
CoT 2026.04	78.69	-
Token-ALP 2026.03	78.1	-
Token-MIS 2026.03	77.84	-
Seq-ALP 2026.03	77.84	-
Seq-MIS 2026.03	77.08	-
Seq-Bypass 2026.03	76.21	-
MCTS + PE 2026.04	76.2	-
GRPO 2026.03	75.91	-
MCTS (Vanilla) 2026.04	75.3	-
Adaptive Parallel Monte Carlo Tree Search (Full Suite) 2026.04	74.8	-
Beam Search 2026.04	72.9	-
FACT-E(Standard) 2026.04	54.09	-
CoT 2026.04	52.18	-
FACT-E(Lightweight) 2026.04	52.18	-
CONSISTENCY 2026.04	51.09	-
REFLECT 2026.04	48.09	-
DENOISE 2026.04	42.62	-
POLISH 2026.04	41	-
BF16 2026.02	0.848	-
HiF4 2026.02	0.826	-2.2
HiF4 2026.02	0.802	5.2
NVFP4 2026.02	0.786	-6.2
NVFP4+PTS 2026.02	0.782	3.2
NVFP4+PTS 2026.02	0.782	-6.6
BF16 2026.02	0.75	-
NVFP4 2026.02	0.74	-1