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

Multimodal Reasoning on WeMath

72.2Accuracy

TGRL-DAPO

Updated 7d ago

Evaluation Results

Method
TGRL-DAPO 2026.03	72.2	-	-	-	-
Perception-R1 2026.03	72	-	-	-	-
TGRL-DAPO 2026.03	71.05	-	-	-	-
Groupwise Ranking Reward 2026.04	70.8	-	-	-	-
NoisyRollout 2026.03	70.3	-	-	-	-
TGRL-GRPO 2026.03	70.29	-	-	-	-
TGRL-GRPO 2026.03	70.23	-	-	-	-
DAPO 2026.03	69.3	-	-	-	-
ThinkLite 2026.03	69.2	-	-	-	-
DAPO 2026.03	68.84	-	-	-	-
GRPO 2026.03	68.2	-	-	-	-
VLAA-Thinker 2026.03	67.9	-	-	-	-
OpenVLThinker 2026.03	67.8	-	-	-	-
GRPO 2026.03	67.4	-	-	-	-
NPO 2026.04	66.95	-	-	-	-
AutoNPO 2026.04	66	-	-	-	-
No Compression 2026.02	63.8	562.6	8.8	-	-
Pointwise GR 2026.04	63.8	-	-	-	-
Prune-on-Logic 2026.02	63.4	419.5	6.6	-	-
Qwen-2.5-VL-7B 2026.03	63.1	-	-	-	-
XMCC 2026.02	63	97.7	1.6	-	-
NPO 2026.04	62.76	-	-	-	-
ExGRPO 2026.04	62.67	-	-	-	-
RLEP 2026.04	62.48	-	-	-	-
SFT-M 2026.02	62.3	-	-	-	-
R1-Onevision 2026.03	62.1	-	-	-	-
StepEntropy 2026.02	61.6	313.5	5.1	-	-
Socratic-Solver-Geo (Stage3) 2026.02	61.58	-	-	-	-
PRM 2026.04	60.8	-	-	-	-
SFT 2026.02	60.23	-	-	-	-
SFT-RS 2026.02	59.2	-	-	-	-
RLVR 2026.04	58.8	-	-	-	-
Athena-PRM 2025.06	58.7	-	-	-	-
TrustGeoGen 2026.02	58.61	-	-	-	-
PGPS9K 2026.02	58.53	-	-	-	-
Socratic-Solver-Geo (Stage2) 2026.02	58.21	-	-	-	-
KD (Our Synthesis) 2026.02	58.15	-	-	-	-
KD (Geo3K) 2026.02	58.02	-	-	-	-
RLSD 2026.04	58	-	-	-	-
GRPO 2026.02	57.99	-	-	-	-
GeoReasoning 2026.02	57.9	-	-	-	-
Qwen2.5-VL-7B-Instruct 2026.02	57.59	-	-	-	-
R-CoT 2026.02	57.59	-	-	-	-
Socratic-Solver-Geo (Stage1) 2026.02	57.54	-	-	-	-
Geo170k (G-LLaVA) 2026.02	57.44	-	-	-	-
Qwen2.5-VL-7B-IT 2026.04	56.6	-	-	-	-
GRPO 2026.04	56.57	-	-	-	-
GRPO 2026.04	56.57	-	-	-	-
No Compression 2026.02	55.8	572.5	10.3	-	-
Athena-ORM 2025.06	55.6	-	-	-	-
GPT-4.1 2025.09	55.5	-	-	-	-
Prune-on-Logic 2026.02	55.4	460.5	8.3	-	-
OPSD 2026.04	54.95	-	-	-	-
XMCC 2026.02	54.9	99.6	1.8	-	-
SFT-M 2026.02	54.89	-	-	-	-
Self-consistency 2025.06	54.8	-	-	-	-
GRPO + OPSD 2026.04	54.76	-	-	-	-
StepEntropy 2026.02	54.6	334.2	6.1	-	-
SFT 2026.02	54.14	-	-	-	-
Base LLM 2026.04	54.1	-	-	-	-
Qwen3-VL-8B-Instruct 2026.04	54.1	-	-	-	-
GRPO 2026.02	52.82	-	-	-	-
SFT-RS 2026.02	52.41	-	-	-	-
LUFFY 2026.04	52.38	-	-	-	-
SDPO 2026.04	52.19	-	-	-	-
Qwen2.5-VL-72B w/ RAPID 2025.06	52.1	-	-	-	-
Ovis2-34B 2025.06	51.9	-	-	-	-
InternVL2.5-38B + VRPRM w/o RL 2025.08	51.43	-	-	-	-
Qwen2.5-VL-32B w/ RAPID 2025.06	50.8	-	-	-	-
SwimBird 2026.02	49.5	-	-	-	-
Claude-3.7-Sonnet 2025.06	49.3	-	-	-	-
Claude3.7-Sonnet 2025.09	49.3	-	-	-	-
VL-Rethinker-72B 2025.06	49.2	-	-	-	-
Qwen2.5-VL-72B 2025.06	49.1	-	-	-	-
Qwen2.5-VL-72B 2025.06	49.1	-	-	-	-
Qwen2.5-VL-72B 2025.06	49.1	-	-	-	-
InternVL2.5-26B + VRPRM w/o RL 2025.08	48.76	-	-	-	-
InternVL3-38B 2025.06	48.6	-	-	-	-
Qwen2.5-VL-32B w/ RAPID 2025.06	48.5	-	-	-	-
PyVision-Image 2026.02	47.7	-	-	-	-
Gemini-2.0-Flash 2025.06	47.4	-	-	-	-
Gemini-2.0-Flash 2025.08	47.4	-	-	-	-
InternVL2.5-38B + VRPRM 2025.08	46.86	-	-	-	-
Athena-PRM 2025.06	46.4	-	-	-	-
InternVL2.5-38B + VisualPRM 2025.08	46.2	-	-	-	-
InternVL3-78B 2025.06	46	-	-	-	-
GPT-4o-20241120 2025.06	45.8	-	-	-	-
GPT-4o 2025.08	45.8	-	-	-	-
LUSPO 2026.02	45.6	-	-	-	-
Qwen2.5-VL-7B w/ RAPID 2025.06	45.4	-	-	-	-
Athena-ORM 2025.06	45.1	-	-	-	-
Ovis2-16B 2025.06	45	-	-	-	-
Self-consistency 2025.06	44.7	-	-	-	-
InternVL2.5-8B + VRPRM w/o RL 2025.08	44.29	-	-	-	-
Claude-3.5-Sonnet 2025.08	44	-	-	-	-
GRPO 2026.02	43.3	-	-	-	-
Qwen2.5-VL-7B (Bo8) 2025.06	43.1	-	-	-	-
Qwen2.5-VL-32B 2025.06	43.1	-	-	-	-
InternVL3-14B 2025.06	43	-	-	-	-
ThinkLite 2025.11	41.8	-	-	-	-

Showing 100 of 191 rows