teacher/v6distill-v1: v6 self-play-imitation distill (UNDERPERFORMED, ladder 1599 ~v5)

teacher/g1: promoted champion (gen1), gate INCONCLUSIVE +39.5 vs gen0 (manual promote)

teacher/g0: 1st self-play champion (gen0), gate PASS +226.8 vs bootstrap, ladder 1664 [1463,1864]

teacher/boot-256x20: VQ-550 Stage-A knowledge warm-start (24.6M, ~v4 strength: top1 39.97%, SF-1600 screen 1345 [1193,1498])

selfplay-best/gen2-g1: browser tfjs bundle (parity-verified)

selfplay-best/gen2-g1: manifest.json

selfplay-best/gen2-g1: sprt.json

selfplay-best/gen2-g1: metrics.json

selfplay-best/gen2-g1: keras_model.keras

selfplay-best/gen2: sprt_ext.log

selfplay-best/gen2: combined_verdict.json

selfplay-best/gen2: manifest.json

selfplay-best/gen2: sprt.json

selfplay-best/gen2: metrics.json

selfplay-best/gen2: keras_model.keras

Archive v6-wdl (experimental WDL value head) + document VQ-498 verdict (strength-neutral, not shipped)

Add v5-vs-v4 held-out confirmation scorecard + value-calibration caveat (VQ-496)

Unify: drop stale root metrics.json (now under v3/)

Unify: drop stale root keras_model.keras (now under v3/)

Unify: all versions (v3/v4/v5) keras+tfjs, Apache-2.0 model card

Upload folder using huggingface_hub

Upload folder using huggingface_hub

initial commit