Prompt Templates Request

[aalmuarik]

Can you please provide the prompt templates used?

In the paper, I could only find the general structure.

1. Issue Type Classification Prompt

[Role & Mission]
Persona setting and Instruction

[Evidence Scope]
Description of the input setting

[Task]
Description the rules the model must follow when responding

[Rules]
Description the rules the model must follow when responding

<Issue Type Set>
["101","102","103","112","Others"]
</Issue Type Set>

<Issue Type Definitions>
Issue Type label Dictionary
</Issue Type Definitions>

[Output Format]
Response Examples with Output Format

---- INPUT ----
<Appellant Arguments>{appellant}</Appellant Arguments>
<Examiner Findings>{examiner}</Examiner Findings>

---

2. Board Authorities Classification Prompt

[Role & Mission]
Persona setting and Instruction

[Evidence Scope]
Description of the input setting

[Task]
Description of the Board Authorities classification task

[Rules]
Description the rules the model must follow when responding
<Board Ruling Dictionary>
[
"37 CFR 1.131",
"37 CFR 1.132",
"37 CFR 41.50",
"37 CFR 41.50(a)",
"37 CFR 41.50(b)",
"37 CFR 41.50(c)",
"37 CFR 41.50(d)",
"37 CFR 41.50(f)",
"Others"
]
</Board Ruling Dictionary>

<Board Ruling Definitions>
Board Authorities label Dictionary
</Board Ruling Definitions>

[Output Format]
Response Examples with Output Format

---- INPUT ----
<Appellant Arguments>{appellant}</Appellant Arguments>
<Examiner Findings>{examiner}</Examiner Findings>

---

3. Subdecision (Fine/Coarse) Classification Prompt

[Role & Mission]
Persona setting and Instruction

[Evidence Scope]
Description of the input setting

[Task]
Description of the Subdecision classification task

[Rules]
Description the rules the model must follow when responding

<Decision Type Dictionary>
fine/coarse subdecision dictionary in the form of {index: label}
</Decision Type Dictionary>

[Output Format]
Response Examples with Output Format

---- INPUT ----
<Appellant Arguments>{appellant}</Appellant Arguments>
<Examiner Findings>{examiner}</Examiner Findings>

[aalmuarik]

Also, which cases where used for evaluations? (15,482)

I was able to align. the ones, that had an application number + >= 2006. 15,478. So only 4 were missing.