Document the Qwen3-VL-8B recaptions folder

README.md

@@ -131,6 +131,21 @@ The two captions of an image share the prompt's fixed sentence order but differ
 
 These are small-model captions. A manual check of 16 captions found the main subject, colors, setting, medium and legible text correct in all of them, but about a third had one wrong secondary detail: left/right or front/back orientation, an object count, or an invented small object. The two captions of an image rarely make the same error. Some interpretive phrasing that is not on the banned lists remains.
 
+
+## Qwen3-VL-8B recaptions
+
+`gemma 3 recaptions/<shard>.parquet` holds one new caption for each of the first 500,000 images (subset `index` 0–499,999). The folder name is the requested upload path. The captions were written by [Qwen/Qwen3-VL-8B-Instruct](https://huggingface.co/Qwen/Qwen3-VL-8B-Instruct), not by Gemma. See `gemma 3 recaptions/README.md` for columns and the generation setup. Join on `key`.
+
+```python
+from datasets import load_dataset
+
+qwen3 = load_dataset(
+    "parquet",
+    data_files="hf://datasets/IntelligenceLab/BLIP3o-Recap/gemma 3 recaptions/*.parquet",
+    split="train",
+)
+```
+
 ## License
 
 Apache-2.0, same as the source dataset.