JGOS-Model/JGOS-31B-Citizen

license: gemma4 파생은 apache-2.0 (google/gemma-4-* 및 Qwen3 모두 apache-2.0)

docs: reposition as administrative/public-sector AI (remove regional specificity)

feat: add Training Datasets section (AIHub K-AI optimized)

docs: rewrite model card in English (professional tone)

docs: enrich tags (darwin-v8, gemma4, k-ai, gpqa, vidraft, etc.)

docs: license gemma, multimodal pipeline_tag, FFN-extract Darwin V8 evolution narrative

docs: K-AI #1 image + Darwin V8 evolution + Korean SFT overview

add K-AI leaderboard #1 screenshot

Update README.md

docs: remove 83.33 from GPQA, keep 84.34 only

eval: remove 83.33 row, keep 84.34 only

Update README.md

add .eval_results/gpqa_diamond.yaml (same structure as Darwin-28B-REASON)

docs: add GPQA Diamond eval (maj@8+DELPHI, near-miss maj)

upload model weights

initial commit