-
v5 model card (100% eval results)
-
v5 SFT dataset (9,923)
-
RemasterPhantom v5 — LoRA adapter
-
RemasterPhantom v5 — merged model (9,923 dedup dataset, real-magic calibration + partial-prefix + variant-select drills)
-
Delete folder lora_adapter with huggingface_hub
-
model card: 칸네리 → 칵네리 표기 통일
-
v4 model card (eval results)
-
v4 SFT dataset (5,306)
-
RemasterPhantom v4 — LoRA adapter
-
RemasterPhantom v4 — merged model (5,306 dedup dataset, variant-contrast + near-miss drills)
-
Delete folder lora_adapter with huggingface_hub
-
v3.1 model card
-
v3.1 SFT dataset (1,941)
-
RemasterPhantom v3.1 — LoRA adapter
-
RemasterPhantom v3.1 — merged model (dedup dataset + full-coverage identify drills + unknown-signature calibration)
-
Delete folder lora_adapter with huggingface_hub
-
SFT dataset
-
RemasterPhantom — LoRA adapter
-
RemasterPhantom — merged model
-
initial commit
