Skip to content

https://arxiv.org/abs/2606.28344. The end of web parsing. The beginning of scalable pixel-native search. link: https://pixelrag.ai/

docs: move Data Curation section from root README to train/README (#59)

Yichuan Wang743a2ff7bad7
diff --git a/README.md b/README.md
index 650cc7f..f35813c 100644
--- a/README.md
+++ b/README.md
@@ -239,13 +239,6 @@ We also release the full training set
([`Chrisyichuan/screenshot-training-natural-filtered-v2`](https://huggingface.co/datasets/Chrisyichuan/screenshot-training-natural-filtered-v2)),
so you can adapt other backbones yourself — a larger Qwen, or any other embedding model.
-### Data Curation
-
-Visualization of some very early version of the training data:
-[early training data viewer](https://yichuan-w.github.io/share/blog-review-first100-light/)
-
-Reproduce: TBD
-
## Acknowledgments
Thanks to [Rulin Shao](https://rulinshao.github.io/) for support.
diff --git a/train/README.md b/train/README.md
index 50cd802..257ff51 100644
--- a/train/README.md
+++ b/train/README.md
@@ -402,3 +402,10 @@ one launch command per run, each adding a single knob. Results are summarized in
For maintainer notes on training internals, hard-negative filtering, dataset
packaging, and tests, see [`docs/training_dev_notes.md`](./docs/training_dev_notes.md).
+
+## Data Curation
+
+Visualization of some very early version of the training data:
+[early training data viewer](https://yichuan-w.github.io/share/blog-review-first100-light/)
+
+Reproduce: TBD