Improve model card for FAAST-Qwen2.5-3B-Instruct

README.md

@@ -1,12 +1,5 @@
 ---
-license: apache-2.0
 base_model: Qwen/Qwen2.5-3B-Instruct
-tags:
-- qwen2.5
-- test-time-learning
-- fast-weights
-- adaptation
-- multilingual
 datasets:
 - OpenWebText2
 - IWSLT2017
@@ -14,25 +7,55 @@ language:
 - en
 - de
 - fr
+license: apache-2.0
+library_name: transformers
+pipeline_tag: text-generation
+tags:
+- qwen2.5
+- test-time-learning
+- fast-weights
+- adaptation
+- multilingual
 ---
 
 # FAAST-Qwen2.5-3B-Instruct
 
-`faast-Qwen2.5-3B-Instruct` is an extension of `Qwen2.5-3B-Instruct` equipped with the FAAST module. The original Qwen2.5-3B-Instruct parameters are frozen, while only the FAAST readout projections are trained.
+`faast-Qwen2.5-3B-Instruct` is an extension of `Qwen2.5-3B-Instruct` equipped with the FAAST module as presented in [FAAST: Forward-Only Associative Learning via Closed-Form Fast Weights for Test-Time Supervised Adaptation](https://huggingface.co/papers/2605.04651). 
 
-The model is designed for efficient test-time learning through fast weights, enabling adaptation without backpropagation (gradient descent).
+The original Qwen2.5-3B-Instruct parameters are frozen, while only the FAAST readout projections are trained. The model is designed for efficient test-time learning through fast weights, enabling adaptation without backpropagation (gradient descent).
+
+The official implementation is available at [https://github.com/baoguangsheng/faast](https://github.com/baoguangsheng/faast).
 
 ## Model Description
 
 FAAST augments Qwen2.5-3B-Instruct with fast-weight adaptation modules that support supervised learning during inference. During FAAST pretraining, all backbone LLM parameters remain frozen, and only lightweight FAAST readout projections are optimized.
 
 This design enables:
-
 - Test-time learning without backpropagation
 - Efficient adaptation with low memory overhead
 - Fast adaptation to downstream tasks
 - Improved few-shot and full-data performance
 
+## Usage
+
+These models can learn at test time. Below is a sample snippet showing how to use the model with `transformers` (requires `trust_remote_code=True`):
+
+```python
+from transformers import AutoTokenizer, AutoModelForCausalLM
+
+model_path = "gshbao/faast-Qwen2.5-3B-Instruct"
+tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
+model = AutoModelForCausalLM.from_pretrained(model_path, trust_remote_code=True)
+
+# Labeled examples for adaptation
+fewshot_samples = ['sample 1', 'sample 2', ...]
+inputs = tokenizer(fewshot_samples, return_tensors="pt", padding=True)
+
+model.reset_projection() # clear existing fast weights
+model.learn(**inputs)    # learn new fast weights from labeled examples
+model.generate(...)      # do the task using the learned fast weights
+```
+
 ## Training Details
 
 - **Base model:** Qwen2.5-3B-Instruct
@@ -82,7 +105,7 @@ BLEU scores on IWSLT2017. Bold scores indicate statistical significance at `p <
 
 ## Citation
 
-If you use this model, please cite the corresponding [FAAST paper](https://arxiv.org/pdf/2605.04651) or [project](https://github.com/baoguangsheng/faast).
+If you use this model, please cite the corresponding [FAAST paper](https://huggingface.co/papers/2605.04651) or [project](https://github.com/baoguangsheng/faast).
 
 ```bibtex
 @article{bao2026faast,