Upload README.md

README.md

@@ -34,7 +34,7 @@ For ONCA 1.5, the training emphasis was to improve all four existing task famili
 | Base model | `Jackrong/Qwen3.5-9B-Claude-4.6-Opus-Reasoning-Distilled` |
 | Architecture | Qwen3.5-class causal LM (`Qwen3_5ForCausalLM`) |
 | Context window | 262,144 tokens |
-| Training recipe | Continued SFT, no DPO |
+| Training recipe | Continued SFT |
 | Domain focus | Pancreatic cancer and oncology-adjacent research workflows |
 
 ## What This Release Is Good At
@@ -44,6 +44,61 @@ For ONCA 1.5, the training emphasis was to improve all four existing task famili
 - Structured pathology abstraction with field-oriented prompting.
 - Variant evidence interpretation with oncology context and uncertainty signaling.
 
+## What It Is Specialized For
+
+This model is specialized for pancreatic cancer and oncology-adjacent research workflows rather than broad general-purpose medical chat. It works best when the task is tightly scoped and the target output format is explicit, especially for:
+
+- pancreatic cancer trial eligibility review
+- pathology report abstraction into structured fields
+- concise oncology reasoning for case discussion
+- variant evidence interpretation with uncertainty signaling
+
+Example: pancreatic cancer trial-screening workflow
+
+```python
+prompt = """
+Task: Pancreatic cancer trial screening.
+
+Patient summary:
+- 63-year-old with metastatic pancreatic ductal adenocarcinoma
+- ECOG 1
+- Prior gemcitabine plus nab-paclitaxel
+- Bilirubin 0.9 mg/dL
+- No active infection
+
+Trial criteria:
+- Histologically confirmed metastatic pancreatic adenocarcinoma
+- ECOG 0-1
+- Progression after 1 prior systemic regimen
+- Adequate marrow and hepatic function
+- Exclude uncontrolled infection
+
+Return:
+1. Eligibility label
+2. Criterion-by-criterion reasoning
+3. Missing information
+"""
+
+messages = [
+    {"role": "system", "content": "You are Onca, a pancreatic cancer clinical research assistant."},
+    {"role": "user", "content": prompt},
+]
+text = tokenizer.apply_chat_template(
+    messages,
+    tokenize=False,
+    add_generation_prompt=True,
+)
+inputs = tokenizer(text, return_tensors="pt").to(model.device)
+outputs = model.generate(
+    **inputs,
+    max_new_tokens=256,
+    do_sample=False,
+    temperature=0.2,
+)
+answer = outputs[0][inputs["input_ids"].shape[1]:]
+print(tokenizer.decode(answer, skip_special_tokens=True))
+```
+
 ## Release Note
 
 This is the main reference checkpoint in the ONCA 1.5 family and the best starting point if you want the least altered release.
@@ -53,7 +108,7 @@ This is the main reference checkpoint in the ONCA 1.5 family and the best starti
 ```python
 from transformers import AutoModelForCausalLM, AutoTokenizer
 
-model_id = "YOUR_USERNAME/onca-1.5"
+model_id = "Joesh1/onca-1.5"
 
 tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
 model = AutoModelForCausalLM.from_pretrained(
@@ -96,7 +151,7 @@ print(tokenizer.decode(answer, skip_special_tokens=True))
 
 ## Training Scope
 
-The active ONCA 1.5 continued-SFT stack uses openly available data only and does not use DPO. The data mixture keeps all four task families in play, with weighting geared toward variant evidence repair and trial-screening stability while guarding against pathology regression.
+The active ONCA 1.5 continued-SFT stack uses openly available data only. The data mixture keeps all four task families in play, with weighting geared toward variant evidence repair and trial-screening stability while guarding against pathology regression.
 
 | Task family | Active rows | Train | Val | Test |
 | --- | ---: | ---: | ---: | ---: |