Update README.md

README.md

@@ -14,26 +14,29 @@ datasets:
 
 # DE-T5-Base-15k
 
-`GermanT5/t5-efficient-gc4-german-base-nl36` continued for **15 000 steps** on the German scientific split (same preprocessing as EN). Checkpoint: `cross_lingual_transfer/logs/native_baseline/.../step-step=015000.ckpt`.
+`GermanT5/t5-efficient-gc4-german-base-nl36` continued for **15 000 steps** on the German portion of the scientific corpus (same preprocessing as EN). Checkpoint: `cross_lingual_transfer/logs/native_baseline/.../step-step=015000.ckpt`.
 
 ## Model Details
-- Base: GermanT5 (T5-base architecture, German tokenizer)
-- Objective: Span corruption (15% noise, mean span length 3)
-- Optimizer: Adafactor, lr=1e-3, warmup 1.5k, inverse sqrt schedule, grad clip 1.0
-- Effective batch: 48
+- Base: GermanT5 (same architecture as T5-base, German tokenizer)
+- Optimizer: Adafactor, lr=1e-3, inverse sqrt schedule, warmup=1.5k, grad clip=1.0
+- Effective batch: 48 (per-GPU 48, no accumulation)
+- Objective: Span corruption (15 % masking, mean span length 3)
 
-## Evaluation (Global-MMLU zero-shot)
+## Training Data
+German split of the Unpaywall-derived corpus (continued-pretraining windows of 512 tokens, 50 % overlap).
+
+## Evaluation (Global-MMLU, zero-shot)
 | Metric | EN | DE |
 | --- | --- | --- |
-| Overall | 0.2295 | 0.2295 |
+| Overall accuracy | 0.2295 | 0.2295 |
 | Humanities | 0.2421 | 0.2421 |
 | STEM | 0.2125 | 0.2125 |
 | Social Sciences | 0.2171 | 0.2171 |
 | Other | 0.2398 | 0.2398 |
 
 ## Intended Use
-Baseline for German scientific text; can be further continued or fine-tuned.
+German scientific NLP baseline; compare against WECHSEL-based models or continue fine-tuning on German datasets.
 
 ## Limitations
-- Modest gains over base GermanT5 (only 15k steps)
-- German SentencePiece vocab; not compatible with English tokenizer.
+- Only 15k steps, so improvements over base GermanT5 are modest.
+- Uses German SentencePiece vocab; incompatible with English tokenizer out of the box.