Qwen3-VL plus DeCoDe lifts accuracy from 27.8% to 89.2% on novel vision tests

HildeKuehne · x · 2026-07-29

Qwen3-VL with DeCoDe jumps from 27.8% to 89.2% on a mixed benchmark suite

The post reports evaluation on 6 standard and 6 newly curated novel benchmarks, covering yoga poses, LEGO bricks, industrial parts, Egyptian hieroglyphs, insects, and Arabic sign language.

With Qwen3-VL + domain information, the DeCoDe method improves performance from 27.8% to 89.2%, and the authors say it outperforms SFT.

The result suggests that domain-aware decomposition can matter a lot when vision-language models face unfamiliar or highly specialized categories.

Original post →

More from Research

• MIT's Neil Thompson: AI comes for tasks before jobs, and fixed costs favor big firms — bigdata · 2026-10-16
• UMD's PAIR framework bridges perception and action in VLA models via bridge tokens — Kaixi Feng · 2026-10-16
• CI-MSE offline metric hits -0.87 Spearman correlation with real robot task success vs weak raw MSE — chris_j_paxton · 2026-10-16
• Colossal Biosciences uses AI to decode genetic switches behind the extinct dodo — castrotech · 2026-10-16
• Colossal Biosciences uses AI to identify gene switches in dodo de-extinction push — castrotech · 2026-10-16
• Cheng Lou open-sources Handwired: hand-written weights turn code into training-free neural nets — dotey · 2026-10-16

Research channel →