List your company
ResearchFormerly WizWand

Multimodal Understanding on MMT emo

60.8AccuracyCoMemo
50.453.155.858.5May 1, 2026
Updated 6mo ago

Evaluation results

MethodLinks
CoMemoTraining Strategy=Visual Injection2026.0560.8
PVM-8B (SFT)Backbone=8B, Training Strategy=SFT, Model Architecture=Persistent Visual Memory2026.0558.3
PVM-8B (SFT + GRPO)Backbone=8B, Training Strategy=SFT + GRPO, Model Architecture=Persistent Visual Memory2026.0558.3
PVM-4B (SFT)Backbone=4B, Training Strategy=SFT, Model Architecture=Persistent Visual Memory2026.0557.5
Qwen3-VL-8B-InstructBackbone=8B2026.0556.7
Qwen3-VL-4B-InstructBackbone=4B2026.0556.7
Qwen3-VL-4B (LoRA-SFT + GRPO)Backbone=4B, Training Strategy=LoRA-SFT + GRPO2026.0555.8
PVM-4B (SFT + GRPO)Backbone=4B, Training Strategy=SFT + GRPO, Model Architecture=Persistent Visual Memory2026.0555.8
PEARL-8BBackbone=8B, Training Strategy=RL-tuned2026.0555
Qwen3-VL-4B (LoRA-SFT)Backbone=4B, Training Strategy=LoRA-SFT2026.0555
MemVRTraining Strategy=Visual Injection2026.0554.2
ICoTTraining Strategy=Visual Injection2026.0554.2
Euclid-8BBackbone=8B, Training Strategy=RL-tuned2026.0554.2
Qwen3-VL-8B (SFT + GRPO)Backbone=8B, Training Strategy=SFT + GRPO2026.0554.2
Qwen3-VL-4B (SFT + GRPO)Backbone=4B, Training Strategy=SFT + GRPO2026.0553.3
Qwen3-VL-8B (LoRA-SFT + GRPO)Backbone=8B, Training Strategy=LoRA-SFT + GRPO2026.0552.5
OneThinker-8BBackbone=8B, Training Strategy=RL-tuned2026.0551.7
Qwen3-VL-8B (LoRA-SFT)Backbone=8B, Training Strategy=LoRA-SFT2026.0551.7
Qwen3-VL-8B (SFT)Backbone=8B, Training Strategy=SFT2026.0550.8
Qwen3-VL-4B (SFT)Backbone=4B, Training Strategy=SFT2026.0550.8