List your company

Multi-modal Image Understanding on LLaVA-Next-7B Evaluation Suite

64.2GQA ScoreVanilla
53.17656.03858.961.762Jun 23, 2026

Evaluation results

MethodLinks
VanillaToken Budget=2880 Tokens, Token Reduction Ratio=100%2026.0664.267.460.61,85186.570.164.957.651.7100
CLSEToken Budget=320 Tokens, Token Reduction Ratio=88.9%2026.0662.667.158.31,79484.467.958.956.141.494.7
HiREDToken Budget=320 Tokens, Token Reduction Ratio=88.9%2026.0659.364.255.91,69083.366.758.854.240.491.7
PDropToken Budget=320 Tokens, Token Reduction Ratio=88.9%2026.0656.463.456.21,66377.667.554.454.125.986.5
SparseVLMToken Budget=320 Tokens, Token Reduction Ratio=88.9%2026.0656.160.654.51,53382.466.158.4522785.8
FastVToken Budget=320 Tokens, Token Reduction Ratio=88.9%2026.0655.961.651.91,66171.762.855.753.137.486.3
PruMergeToken Budget=320 Tokens, Token Reduction Ratio=88.9%2026.0653.661.355.31,53460.866.450.65414.679.2