List your company

Multimodal Reasoning on ScienceQA

97.85Average AccuracyMG2-RAG
71.48678.330585.17592.0195Nov 4, 2025Dec 5, 2025Jan 6, 2026Feb 6, 2026Mar 10, 2026Apr 10, 2026May 12, 2026

Evaluation results

MethodLinks
MG2-RAGLLM/MLLM=Qwen3.5-27B2026.0497.8598.4595.2898.7398.4496.4398.6898.4296.84—
Qwen3.5-27BLLM/MLLM=Qwen3.5-27B2026.0497.3498.2795.1697.1897.996.4397.4997.7696.57—
MG2-RAGLLM/MLLM=GPT-5.22026.049797.7894.6597.0697.9396.2497.0597.3396.43—
GPT-5.2LLM/MLLM=GPT-5.22026.0496.397.693.0595.9697.7395.3596.1796.5495.88—
Gemini-3.1-ProLLM/MLLM=Gemini-3.1-Pro2026.0492.993.5394.1290.8192.9894.2590.2794.0390.93—
MG2-RAGLLM/MLLM=Qwen2.5-VL-7B2026.0489.2590.8187.9687.0989.9386.6188.9290.6486.75—
HumanLLM/MLLM=-2026.0488.490.2384.9787.4889.687.588.191.5982.42—
MG2-RAGLLM/MLLM=Qwen2.5-72B2026.0487.6288.3781.2191.278778.2894.0188.8485.43—
VISTA-SFTModel Variant=Qwen2.5-VL-7B2026.0587.58—————————
Qwen2.5-VL-7BLLM/MLLM=Qwen2.5-VL-7B2026.0487.588.7788.8683.8287.3988.0586.0689.9483.12—
LaReBase VLM=Qwen3-VL-4B-Instruct2025.1187.5————————65.2
R3VModel Variant=Qwen2.5-VL-7B2026.0586.83—————————
STaR+Model Variant=Qwen2.5-VL-7B2026.0585.73—————————
SFT-OracleModel Variant=Qwen2.5-VL-7B2026.0585.49—————————
VaLiKLLM/MLLM=Qwen2.5-72B2026.0484.7785.6175.9390.2784.474.1792.3385.7982.98—
MG2-RAGLLM/MLLM=Qwen2.5-7B2026.0484.5385.1781.3385.8283.4376.189.4186.1281.67—
STaRModel Variant=Qwen2.5-VL-7B2026.0584.39—————————
CoTLLM/MLLM=GPT-42026.0483.9985.4872.4490.2782.6571.4992.8986.6679.04—
CCoTBase VLM=Qwen3-VL-4B-Instruct2025.1183.4————————240.2
VaLiKLLM/MLLM=Qwen2.5-7B2026.0483.1684.1575.1487.6482.9973.1889.6984.480.95—
MM-CoTBase VLM=Qwen3-VL-4B-Instruct2025.1182.7————————203.1
LaReBase VLM=Qwen3-VL-2B-Instruct2025.1182.5————————55
VISTA-SFTModel Variant=Qwen2.5-VL-3B2026.0582.29—————————
SFT-OracleModel Variant=Qwen2.5-VL-3B2026.0582—————————
ICoTBase VLM=Qwen3-VL-4B-Instruct2025.1181.7————————172.6
SFT-SeedModel Variant=Qwen2.5-VL-7B2026.0581.55—————————
R3VModel Variant=Qwen2.5-VL-3B2026.0580.81—————————
CoF*Base VLM=Qwen3-VL-4B-Instruct2025.1180.6————————205.5
LVR*Base VLM=Qwen3-VL-4B-Instruct2025.1180.3————————101.4
MM-CoTBase VLM=Qwen3-VL-2B-Instruct2025.1180.1————————195.7
ChameleonLLM/MLLM=ChatGPT2026.0479.9381.6270.648479.7770.886.6281.8676.53—
CCoTBase VLM=Qwen3-VL-2B-Instruct2025.1179.9————————200.9
Standard CoTBase VLM=Qwen3-VL-4B-Instruct2025.1179.5————————186.2
LVR*Base VLM=Qwen3-VL-2B-Instruct2025.1179.2————————93.8
MCOUTBase VLM=Qwen3-VL-4B-Instruct2025.1178.8————————97.5
STaR+Model Variant=Qwen2.5-VL-3B2026.0578.6—————————
MMGraphRAGLLM/MLLM=Qwen2.5-VL-7B2026.0478.2181.0868.6280.0979.5268.9683.6980.8773.43—
STaRModel Variant=Qwen2.5-VL-3B2026.0577.49—————————
ICoTBase VLM=Qwen3-VL-2B-Instruct2025.1177————————174.2
MCOUTBase VLM=Qwen3-VL-2B-Instruct2025.1176.8————————104.3
Visual GenomeLLM/MLLM=Qwen2.5-7B2026.0475.0876.7867.0478.0974.0566.1979.7278.08——
SFT-SeedModel Variant=Qwen2.5-VL-3B2026.0574.66—————————
Standard CoTBase VLM=Qwen3-VL-2B-Instruct2025.1174.4————————182.5
MMKGLLM/MLLM=Qwen2.5-7B2026.0473.4773.9866.3778.1871.6564.379.6576.5168.03—
CoF*Base VLM=Qwen3-VL-2B-Instruct2025.1172.5————————195.7