List your company
ResearchFormerly WizWand

Instruction Following on ALFRED

19.84AccuracyTrimTokenator-LC
9.419212.124614.8317.5354Dec 28, 2025Jan 14, 2026Feb 1, 2026Feb 19, 2026Mar 9, 2026Mar 27, 2026Apr 14, 2026
Updated 6mo ago

Evaluation results

MethodLinks
TrimTokenator-LCBackbone=Yi-VL-6B, Retention Ratio=0.22025.1219.84
DivPruneBackbone=Yi-VL-6B, Retention Ratio=0.22025.1218.76
TrimTokenatorBackbone=Yi-VL-6B, Retention Ratio=0.22025.1218.67
Original ModelBackbone=LLaVA-1.5-13B, Retention Ratio=1.02025.1218.64
SparseVLMBackbone=Yi-VL-6B, Retention Ratio=0.22025.1218.35
VisionZipBackbone=Yi-VL-6B, Retention Ratio=0.22025.1218.17
TrimTokenator-LCBackbone=LLaVA-1.5-13B, Retention Ratio=0.22025.1218.14
VTWBackbone=Yi-VL-6B, Retention Ratio=0.22025.1217.87
CATPBackbone=Yi-VL-6B, Retention Ratio=0.22025.1217.82
FastVBackbone=Yi-VL-6B, Retention Ratio=0.22025.1217.62
Full CacheArchitecture=Qwen3-VL, Parameter Scale=8B2026.0416.62
H2OArchitecture=Qwen3-VL, Parameter Scale=8B2026.0416.36
RetentiveKVArchitecture=Qwen3-VL, Parameter Scale=8B2026.0416.34
Full CacheArchitecture=LLaVA-v1.5, Parameter Scale=7B2026.0416.32
TrimTokenatorBackbone=LLaVA-1.5-13B, Retention Ratio=0.22025.1216.25
LOOK-MArchitecture=Qwen3-VL, Parameter Scale=8B2026.0416.22
SparseVLMBackbone=LLaVA-1.5-13B, Retention Ratio=0.22025.1216.21
Original ModelBase Model=InternVL-Chat-ViT-6B-Vicuna-13B2025.1216.2
SAINTArchitecture=Qwen3-VL, Parameter Scale=8B2026.0416.15
SnapKVArchitecture=Qwen3-VL, Parameter Scale=8B2026.0416.12
TrimTokenator-LCBase Model=InternVL-Chat-ViT-6B-Vicuna-13B2025.1216.1
Original ModelBackbone=Yi-VL-6B, Retention Ratio=1.02025.1216.02
MedaArchitecture=Qwen3-VL, Parameter Scale=8B2026.0415.98
Full CacheArchitecture=Qwen3-VL, Parameter Scale=4B2026.0415.98
RetentiveKVArchitecture=Qwen3-VL, Parameter Scale=4B2026.0415.87
RetentiveKVArchitecture=LLaVA-v1.5, Parameter Scale=7B2026.0415.82
VisionZipBackbone=LLaVA-1.5-13B, Retention Ratio=0.22025.1215.76
SnapKVArchitecture=Qwen3-VL, Parameter Scale=4B2026.0415.64
CATPBackbone=LLaVA-1.5-13B, Retention Ratio=0.22025.1215.62
DivPruneBackbone=LLaVA-1.5-13B, Retention Ratio=0.22025.1215.5
VTWBackbone=LLaVA-1.5-13B, Retention Ratio=0.22025.1215.35
LOOK-MArchitecture=Qwen3-VL, Parameter Scale=4B2026.0415.32
MedaArchitecture=Qwen3-VL, Parameter Scale=4B2026.0415.26
DivPruneBase Model=InternVL-Chat-ViT-6B-Vicuna-13B2025.1215.23
SAINTArchitecture=Qwen3-VL, Parameter Scale=4B2026.0415.19
FastVBackbone=LLaVA-1.5-13B, Retention Ratio=0.22025.1215.17
TrimTokenatorBase Model=InternVL-Chat-ViT-6B-Vicuna-13B2025.1215.12
SparseVLMBase Model=InternVL-Chat-ViT-6B-Vicuna-13B2025.1214.94
MedaArchitecture=LLaVA-v1.5, Parameter Scale=7B2026.0414.87
Original ModelBackbone=LLaVA-1.5-7B, Retention Ratio=1.02025.1214.81
VisionZipBase Model=InternVL-Chat-ViT-6B-Vicuna-13B2025.1214.76
H2OArchitecture=Qwen3-VL, Parameter Scale=4B2026.0414.67
SAINTArchitecture=LLaVA-v1.5, Parameter Scale=7B2026.0414.64
SnapKVArchitecture=LLaVA-v1.5, Parameter Scale=7B2026.0414.57
H2OArchitecture=LLaVA-v1.5, Parameter Scale=7B2026.0414.49
LOOK-MArchitecture=LLaVA-v1.5, Parameter Scale=7B2026.0414.33
CATPBase Model=InternVL-Chat-ViT-6B-Vicuna-13B2025.1214.15
FastVBase Model=InternVL-Chat-ViT-6B-Vicuna-13B2025.1213.73
VTWBase Model=InternVL-Chat-ViT-6B-Vicuna-13B2025.1213.62
TrimTokenator-LCBackbone=LLaVA-1.5-7B, Retention Ratio=0.22025.1212.74
DivPruneBackbone=LLaVA-1.5-7B, Retention Ratio=0.22025.1211.31
TrimTokenatorBackbone=LLaVA-1.5-7B, Retention Ratio=0.22025.1211.26
VisionZipBackbone=LLaVA-1.5-7B, Retention Ratio=0.22025.1211.18
SparseVLMBackbone=LLaVA-1.5-7B, Retention Ratio=0.22025.1211.06
CATPBackbone=LLaVA-1.5-7B, Retention Ratio=0.22025.1210.32
VTWBackbone=LLaVA-1.5-7B, Retention Ratio=0.22025.1210.28
FastVBackbone=LLaVA-1.5-7B, Retention Ratio=0.22025.129.82