amad-iq/amad-vlm6-technical-report

Revise after external review: corrected WER/DIMI/lineage claims, behavior stats, full decomposition, bootstrap CIs, cap analysis, missing-control section

Rename merged model to amad-vlm6; link released weights

Limitations: single-benchmark scope — unseen pros and cons outside KITAB-Bench

Add author attribution: Murtadha (github.com/murtadha-lap)

Enhanced report: full 13-dataset results, error-type decomposition, practitioner lessons, reproduction recipe

amad-vlm5-D technical report: TIES merge of amad-vlm5 and DIMI-Arabic-OCR-V2

initial commit