[codex] add neural watermark scaffold#1

Merged

CEQ151 merged 32 commits into

masterfrom

codex/mlwm-v1

Apr 27, 2026

Commits on Apr 26, 2026

feat(mlwm): add neural watermark scaffold

CEQ151
authored and
CEQ151
committed
fix(ci): update watermark smoke workflow

CEQ151
authored and
CEQ151
committed
ci: expand mlwm validation checks

CEQ151
authored and
CEQ151
committed
fix(mlwm): harden training export smoke path

CEQ151
authored and
CEQ151
committed
feat(mlwm): add dataset preparation helper

CEQ151
authored and
CEQ151
committed
feat(mlwm): support unsplash lite ingestion

CEQ151
authored and
CEQ151
committed
docs: add branch wiki documentation

CEQ151
authored and
CEQ151
committed
docs: internationalize project overview

CEQ151
authored and
CEQ151
committed
docs: split english and chinese readmes

CEQ151
authored and
CEQ151
committed
feat(mlwm): add training dashboard

CEQ151
authored and
CEQ151
committed
feat(mlwm): stream live training telemetry

CEQ151
authored and
CEQ151
committed
feat(mlwm): add cloud training configs

CEQ151
authored and
CEQ151
committed
fix(mlwm): whiten neural payload bits

CEQ151
authored and
CEQ151
committed
fix(mlwm): preserve decoder spatial payload grid

CEQ151
authored and
CEQ151
committed
feat(mlwm-train): add clean encoder bootstrap config

CEQ151
authored and
CEQ151
committed
fix(mlwm-train): keep clean bootstrap differentiable

CEQ151
authored and
CEQ151
committed
fix(mlwm-train): guide encoder bootstrap with teacher residual

CEQ151
authored and
CEQ151
committed
fix(mlwm-train): add staged encoder curriculum

CEQ151
authored and
CEQ151
committed
fix(mlwm): inject spatial payload grid into encoder

CEQ151
authored and
CEQ151
committed
feat(mlwm-train): resume fine-tuning from checkpoint

CEQ151
authored and
CEQ151
committed
feat(mlwm-train): track RS decode success

CEQ151
authored and
CEQ151
committed
feat(mlwm): add checkpoint evaluator

CEQ151
authored and
CEQ151
committed
feat(mlwm-train): add boosted robustness fine-tune

CEQ151
authored and
CEQ151
committed
feat(mlwm-train): add crop-focused attack curriculum

CEQ151
authored and
CEQ151
committed
fix(mlwm-train): train decoder on encoder residuals

CEQ151
authored and
CEQ151
committed
feat(mlwm-train): add main-only decode candidate config

CEQ151
authored and
CEQ151
committed

Commits on Apr 27, 2026

fix(mlwm-export): preserve trained residual scale

CEQ151
authored and
CEQ151
committed
fix(mlwm-export): use onnx-friendly image size

CEQ151
authored and
CEQ151
committed
feat(mlwm): promote alpha1 neural model

CEQ151
authored and
CEQ151
committed
test(mlwm): avoid training dependency in runtime attack tests

CEQ151
authored and
CEQ151
committed
chore(mlwm): archive alpha1 training checkpoint

CEQ151
authored and
CEQ151
committed
Merge remote-tracking branch 'origin/master' into codex-mlwm-v1

CEQ151
authored and
CEQ151
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[codex] add neural watermark scaffold#1

[codex] add neural watermark scaffold#1
CEQ151 merged 32 commits into
masterfrom
codex/mlwm-v1

Commits on Apr 26, 2026

feat(mlwm): add neural watermark scaffold

fix(ci): update watermark smoke workflow

ci: expand mlwm validation checks

fix(mlwm): harden training export smoke path

feat(mlwm): add dataset preparation helper

feat(mlwm): support unsplash lite ingestion

docs: add branch wiki documentation

docs: internationalize project overview

docs: split english and chinese readmes

feat(mlwm): add training dashboard

feat(mlwm): stream live training telemetry

feat(mlwm): add cloud training configs

fix(mlwm): whiten neural payload bits

fix(mlwm): preserve decoder spatial payload grid

feat(mlwm-train): add clean encoder bootstrap config

fix(mlwm-train): keep clean bootstrap differentiable

fix(mlwm-train): guide encoder bootstrap with teacher residual

fix(mlwm-train): add staged encoder curriculum

fix(mlwm): inject spatial payload grid into encoder

feat(mlwm-train): resume fine-tuning from checkpoint

feat(mlwm-train): track RS decode success

feat(mlwm): add checkpoint evaluator

feat(mlwm-train): add boosted robustness fine-tune

feat(mlwm-train): add crop-focused attack curriculum

fix(mlwm-train): train decoder on encoder residuals

feat(mlwm-train): add main-only decode candidate config

Commits on Apr 27, 2026

fix(mlwm-export): preserve trained residual scale

fix(mlwm-export): use onnx-friendly image size

feat(mlwm): promote alpha1 neural model

test(mlwm): avoid training dependency in runtime attack tests

chore(mlwm): archive alpha1 training checkpoint

Merge remote-tracking branch 'origin/master' into codex-mlwm-v1