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

Text-based Visual Question Answering on TextVQA (ANLS)

60.7ANLS

HONES (tuned)

Updated 8d ago

Evaluation Results

Method	Links
HONES (tuned) 2026.04		60.7
HONES (zero-shot) 2026.04		60.1
Base 2026.04		59.6
RandNeuron 2026.04		59.6
HONES (tuned) 2026.04		48.8
HONES (zero-shot) 2026.04		47.4
Base 2026.04		47
RandNeuron 2026.04		46.9
Qwen2-VL-2B 2026.05		0.799
MiniCPM-V2 2026.05		0.743
VEN-VL 2026.05		0.737
InternVL2-2B 2026.05		0.735
VEN-VL 2026.05		0.655
Qwen-VL-Chat 2026.05		0.615
VEN-VL 2026.05		0.61
LLaVA-MOD 2026.05		0.599
LLaVA-Mini 2026.05		0.591
LLaVA-MOD 2026.05		0.585
LLaVA-1.5 2026.05		0.582
TinyLLaVA† 2026.05		0.574
PruMerge+ 2026.05		0.571
Mini-Gemini-2B 2026.05		0.562
PruMerge 2026.05		0.56
VisionZip 2026.05		0.555
Imp-2B 2026.05		0.545
Bunny-2B 2026.05		0.532
SparseVLM 2026.05		0.518
PyramidDrop 2026.05		0.506
TinyLLaVA† 2026.05		0.492
MoE-LLaVA-2B 2026.05		0.48
FastV 2026.05		0.478
TinyLLaVA† 2026.05		0.477
TinyLLaVA† 2026.05		0.474