Update README.md

README.md

@@ -17,8 +17,7 @@ tags:
 # African Languages Sentiment Classifier (Hausa, Yorùbá, Swahili)
 
 A LoRA-adapted sentiment classifier for Hausa, Yorùbá, and Swahili, fine-tuned
-on [`gospelgit/African-Languages_Sentiments`](https://huggingface.co/datasets/gospelgit/African-Languages_Sentiments)
-— a combined dataset of **46,725 rows** stitched from three independent
+on a combined dataset of **46,725 rows** stitched from three independent
 sources across three different domains, built to reduce the single-domain
 (Twitter-only) bias common in existing African-language sentiment resources.
 
@@ -34,9 +33,7 @@ sources across three different domains, built to reduce the single-domain
 
 ## Training Data
 
-The full dataset card, source breakdown, and licensing details live at
-[`gospelgit/African-Languages_Sentiments`](https://huggingface.co/datasets/gospelgit/African-Languages_Sentiments).
-Summary:
+This model was trained on a combined dataset built from three sources:
 
 | Source | Domain | Languages | Rows |
 |---|---|---|---|
@@ -47,16 +44,14 @@ Summary:
 3-class labels (`positive` / `negative` / `neutral`), 70/15/15 train/dev/test
 split per language, stratified by label.
 
-> **Note on training data adaptation**: this specific adapter was trained on
-> an AutoScientist-evolved version of the dataset above — its "Adaptive
-> Data" step rewrote the original rows into `enhanced_prompt`/
-> `enhanced_completion` pairs (15,280 rows after this process) as part of
-> its data-and-recipe co-optimization loop. Both versions are available in
-> the [`gospelgit/African-Languages_Sentiments`](https://huggingface.co/datasets/gospelgit/African-Languages_Sentiments)
-> repo: the original combined dataset described above, and the
-> AutoScientist-adapted version this model was actually trained on. If you
-> want the raw, unmodified rows for your own training pipeline, use the
-> original files rather than the adapted ones.
+> **Note on training data adaptation**: this adapter was trained on a
+> version of this data that was adapted once via [Adaption Labs'
+> AutoScientist](https://adaptionlabs.ai/blog/autoscientist-challenge) —
+> its "Adaptive Data" step rewrote the original rows into
+> `enhanced_prompt`/`enhanced_completion` pairs (15,280 rows after this
+> process) as part of its data-and-recipe co-optimization loop. **The
+> dataset used to produce this result is included in this repo, alongside
+> the model weights** (see Files and versions).
 
 ## Training Procedure