Update README.md

README.md

@@ -5,6 +5,5 @@ pipeline_tag: zero-shot-classification
 
 # Model Card for Model ID
 
-Continuous fine-tuning of deberta-tasksource, fine-tuned on newer tasksource and with context length of size 1024.
-
-Upcoming: longer training + 1280 tokens context length.
+deberta-v3-base with context length of 1280 fine-tuned on tasksource for 150k steps. I oversampled tasks like ConTRoL, doc-nli.
+Training data include helpsteer v1/v2, logical reasoning tasks (FOLIO, FOL-nli, LogicNLI...), OASST, hh/rlhf, linguistics oriented NLI tasks, tasksource-dpo, fact verification tasks.