List your company

Math on AIME 2025 (accuracy (%))

69.9Accuracy (%)VP2O
-2.79616.07734.9553.823Apr 6, 2026Apr 16, 2026Apr 26, 2026May 6, 2026May 16, 2026May 26, 2026Jun 6, 2026

Evaluation results

MethodLinks
VP2OContext Length=16K2026.0669.9
GRPO-baselineContext Length=16K2026.0667.8
LLaDA-2.1-flashParams=100B-A5B, Sampling mode=Diffusion2026.0663.33
VP2OContext Length=8K2026.0662.5
LLaDA-2.0-flashParams=100B-A5B, Sampling mode=Diffusion2026.0660
Qwen3.5-9BParams=9B, Sampling mode=AR2026.0660
GRPO-baselineContext Length=8K2026.0659.6
FLARE-9BParams=9B, Sampling mode=AR-Trust2026.0654.44
FLARE-9BParams=9B, Sampling mode=Diffusion-Trust2026.0653.33
LLaDA-2.0-miniParams=16B-A1B, Sampling mode=Diffusion2026.0636.67
LLaDA-2.1-miniParams=16B-A1B, Sampling mode=Diffusion2026.0636.67
Original (T2T)Backbone=LLaDA2.1-mini, Inference Strategy=Text-to-Text2026.0433.33
T2MBackbone=LLaDA2.1-mini, Inference Strategy=Token-to-Mask, Remasking Strategy=LOWPROB, τ=0.3, Cmax=1, ρmax=0.252026.0433.33
TIESBackbone=Qwen2.5-7B-Base2026.0620
RAMBackbone=Qwen2.5-7B-Base2026.0616.67
RESMERGEBackbone=Qwen2.5-7B-Base, configuration=SRC-A+LHC2026.0616.67
DARE + TIESBackbone=Qwen2.5-7B-Base2026.0613.33
TSV-MergeBackbone=Qwen2.5-7B-Base2026.0613.33
SDAR 30B-A3BParams=30B-A3B, Sampling mode=Diffusion2026.0610.8
SimpleRLBackbone=Qwen2.5-7B-Base2026.0610
ZeroBackbone=Qwen2.5-7B-Base2026.0610
ReasonerBackbone=Qwen2.5-7B-Base2026.0610
ISO-CBackbone=Qwen2.5-7B-Base2026.0610
ISO-CTSBackbone=Qwen2.5-7B-Base2026.0610
TABackbone=Qwen2.5-7B-Base2026.063.33
Qwen2.5-7BBackbone=Qwen2.5-7B-Base2026.060