File size: 1,012 Bytes
3a81d1e
 
5699b6e
3a81d1e
 
 
 
 
 
 
 
 
 
 
 
 
5699b6e
d93be1c
 
3a81d1e
 
552f723
 
 
 
 
 
3a81d1e
552f723
 
 
 
 
 
3a81d1e
552f723
3a81d1e
d93be1c
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
name: Evasion Bench
description: >
  EvasionBench is a benchmark dataset for detecting evasive answers in earnings
  call Q&A sessions. The task is to classify how directly corporate management
  addresses questions from financial analysts.
tasks:
  - id: evasion_bench
    config: default
    split: train

    epochs: 1

    shuffle_choices: true

    field_spec:
      input: question
      target: "eva4b_label_letter"
      choices: "choices"
      metadata: ["answer"]

    solvers:
      - name: prompt_template
        args:
          template: |
            Question: {prompt}

            Answer: {answer}
      - name: multiple_choice
        args:
          template: |
            You are a financial analyst. Your task is to Detect Evasive Answers in Financial Q&A.
            The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of {letters}.

            {question}

            {choices}
    scorers:
      - name: choice