Table 1 Instruction tuning examples for oncology tasks.

Hallmarks of Cancer (HoC)

As a medical expert, assess the clinical text for cancer hallmarks. Assign one or more labels from the list: Sustaining proliferative signaling (PS), Enabling replicative immortality (RI), Inducing angiogenesis (A), Genome instability & mutation (GI), Tumor-promoting inflammation (TPI), ...

Taken together, the present study clearly shows the synergistic anti-inflammatory as well as anti-oxidative stress effects of CUR and PUFA.

Tumor-promoting inflammation (TPI)

Natural Language Inference (MedNLI)

Evaluate the connection between two clinical sentences and classify them into one of these categories: Contradiction (if the sentences conflict), Neutral (if no logical association), or Entailment (if one sentence logically implies the other)...

Sentence 1: Lung cancer as above s/p pneumonectomy Sentence 2: History of smoking.

Neutral

Relationship Extraction (i2b2-2010)

In the clinical text, your objective is to identify relationships between medical problems, treatments, and tests. Medical problems are tagged as @problem$, medical tests as @test$, and treatments as @treatment$. Classify the relationship as: Treatment is administered for medical problem (TrAP)...

His past medical history is significant for prostate cancer, benign prostatic hypertrophy, hypothyroidism, status post @treatment$ for @problem$, chronic painless hematuria, degenerative joint disease, and history of a murmur.

TrAP

Named Entity Recognition (NER)

Your mission is to tag disease-related Named Entities in the text using the BIO labeling scheme. When you encounter a disease-related phrase, mark the start with B (Begin) and continue with I (Inner) ...

Its role in the therapy of glomerulonephritis, autoimmunity, cystic renal diseases and renal cancer is under investigation.

... cystic: B, renal: I, diseases: I, and: O, renal: B, cancer: I...