Update README.md

README.md

@@ -122,6 +122,158 @@ The model is designed as a research tool for factor selection and should be used
 
 ---
 
+# Usage
+
+Alpha-R1 is designed for **alpha screening** in quantitative investment research rather than general-purpose conversation.
+
+Given:
+
+- Current market conditions
+- Historical market memory
+- Candidate alpha factor descriptions
+- Asset universe information
+
+the model reasons about factor effectiveness under the prevailing market regime and selects factors that are more likely to generate excess returns.
+
+## Example
+
+```python
+from transformers import AutoTokenizer, AutoModelForCausalLM
+import torch
+
+model_name = "FinStep-AI/Alpha-R1"
+
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+
+model = AutoModelForCausalLM.from_pretrained(
+    model_name,
+    torch_dtype="auto",
+    device_map="auto"
+)
+
+model.eval()
+
+system_prompt = """
+You are a senior quantitative investment expert, skilled in selecting
+the most suitable alpha factor combinations based on market environment
+and asset characteristics.
+
+You need to analyze current market conditions, the characteristics of
+each factor, and asset portfolio situations to provide scientific and
+reasonable factor selection recommendations.
+"""
+
+user_prompt = """
+Based on the following information, select the most suitable factor
+combinations for {target_date}'s trading day for a {holding_days}-day
+short-term strategy stock selection
+(buy at market open, sell at market close after {holding_days} trading days).
+
+Target Date:
+{target_date}
+
+Market Environment Information:
+
+• Previous Trading Day Closing Data ({previous_trading_day}):
+{market_price_data}
+
+• Previous Trading Day Market Analysis ({previous_trading_day}):
+{market_analysis}
+
+• Current Day Pre-Market News ({target_date}):
+{financial_news}
+
+Available Factor Descriptions:
+
+• {factor_1_name}:
+{factor_1_description}
+
+• {factor_2_name}:
+{factor_2_description}
+
+• {factor_3_name}:
+{factor_3_description}
+
+...
+
+Asset Portfolio Information:
+
+{asset_pool_information}
+
+Analysis Framework:
+
+1. Analyze each factor's nature and characteristics.
+2. Evaluate factors' expected performance in the current market.
+3. Consider portfolio characteristics for further screening.
+4. Make the final selection (maximum 10 factors).
+
+Output Requirements:
+
+• Provide detailed analytical reasoning first.
+• Output XML-tagged factor list:
+  <alpha_list><alpha001>...</alpha_list>
+• Maximum 10 factors allowed.
+• Skip selection if no factors are expected to yield positive returns.
+"""
+
+messages = [
+    {"role": "system", "content": system_prompt},
+    {"role": "user", "content": user_prompt},
+]
+
+text = tokenizer.apply_chat_template(
+    messages,
+    tokenize=False,
+    add_generation_prompt=True,
+)
+
+inputs = tokenizer(
+    text,
+    return_tensors="pt"
+).to(model.device)
+
+with torch.no_grad():
+    outputs = model.generate(
+        **inputs,
+        max_new_tokens=4096,
+        temperature=0.6,
+        top_p=0.95,
+        pad_token_id=tokenizer.eos_token_id,
+    )
+
+response = tokenizer.decode(
+    outputs[0][inputs.input_ids.shape[1]:],
+    skip_special_tokens=True,
+)
+
+print(response)
+```
+
+## Expected Output
+
+```text
+Factor analysis reasoning: [Detailed explanation of selection logic...]
+The most suitable factor selection for the current market is: <alpha_list><alpha001><alpha003><alpha007></alpha_list>
+```
+
+## Generation Recommendations
+
+```python
+generation_config = {
+    "temperature": 0.6,
+    "top_p": 0.95,
+    "max_new_tokens": 4096,
+}
+```
+
+For reproducible factor-screening results, we recommend:
+
+```python
+temperature=0
+```
+
+which is also consistent with the evaluation setting reported in the paper.
+
 # Limitations
 
 - This model is **not** a financial advisor.
@@ -136,7 +288,7 @@ The model is designed as a research tool for factor selection and should be used
 If you use Alpha-R1 in your research, please cite our paper:
 
 ```bibtex
-@article{alphar1,
+@article{jiang2025alphar1,
   title={Alpha-R1: Alpha Screening with LLM Reasoning via Reinforcement Learning},
   author={Jiang, Zuoyou and Zhao, Li and Sun, Rui and Sun, Ruohan and Li, Zhongjian and Li, Jing and Jiang, Daxin and Bai, Zuo and Hua, Cheng},
   journal={arXiv preprint arXiv:2512.23515},