Card: point to the paper and configs for training settings; amplification is not part of COUNT

README.md

@@ -19,8 +19,12 @@ Training-Free Adaptation** ([paper](https://arxiv.org/abs/XXXX.XXXXX), [code](ht
 COUNT adds an n-gram embedding to a standard transformer without adding learned parameters. For each
 position it looks up the most frequent next tokens of the longest matching n-grams in a table of corpus
 counts held in CPU RAM, averages those tokens' rows of the model's own output head (weighted by their
-count scores), and adds the result to the token embedding. A LayerNorm and a √d scale follow
-(embedding amplification), whose two small parameter vectors are the only addition to the backbone.
+count scores), and adds the result to the token embedding.
+
+These checkpoints were also trained with other settings that are not part of COUNT itself, such as
+embedding amplification. For the full configuration of each model, see the paper's appendix, the
+`config.json` and `release_args.json` in each step directory, and the ladder configuration in the
+[code](https://github.com/IanMagnusson/COUNT).
 
 ## Checkpoints
 
@@ -41,8 +45,7 @@ weights, optimizer state and trainer state.
 
 Paths follow the code's ladder layout, `model-ladders/count-<size>/<SIZE>/step<N>/`, so pointing
 `OLMO_ROOT_DIR` at a download lets the code find them. Each step directory has a `release_args.json`
-with the ladder settings it was trained with: COUNT reads 8 of the table's 32 candidates per order
-(`--count-k 8`) and averages over orders 1–6 (`--count-max-order 6 --count-num-orders 6`).
+with the ladder settings it was trained with.
 
 The tables are in [count-ngram-tables](https://huggingface.co/datasets/IanMagnusson/count-ngram-tables).
 A checkpoint can be evaluated with its training table or with any other table there, including the