DrivelologyBinaryClassification#

Overview#

Drivelology Binary Classification evaluates models’ ability to identify “drivelology” - a unique linguistic phenomenon characterized as “nonsense with depth.” These are utterances that are syntactically coherent yet pragmatically paradoxical, emotionally loaded, or rhetorically subversive.

Task Description#

  • Task Type: Binary Text Classification (Yes/No)

  • Input: Text sample to classify

  • Output: “Yes” if drivelology, “No” otherwise

  • Domain: Linguistic analysis, humor detection, pragmatics

Key Features#

  • Tests understanding of layered linguistic meanings

  • Distinguishes nonsense-with-depth from pure nonsense and normal text

  • Requires contextual understanding and emotional insight

  • Covers humor, irony, sarcasm detection

  • Multiple difficulty levels available

Evaluation Notes#

  • Default configuration uses 0-shot evaluation

  • Metrics: Accuracy, Precision, Recall, F1-Score

  • Subsets: binary-english-easy, binary-english-hard, binary-chinese-easy, binary-chinese-hard

Properties#

Property

Value

Benchmark Name

drivel_binary

Dataset ID

extraordinarylab/drivel-hub

Paper

N/A

Tags

Yes/No

Metrics

accuracy, precision, recall, f1_score, yes_ratio

Default Shots

0-shot

Evaluation Split

test

Aggregation

f1

Data Statistics#

Metric

Value

Total Samples

1,200

Prompt Length (Mean)

1056.08 chars

Prompt Length (Min/Max)

984 / 1449 chars

Sample Example#

Subset: binary-classification

{
  "input": [
    {
      "id": "ddbda8da",
      "content": [
        {
          "text": "#Instruction#:\nClassify whether the given text is a Drivelology sample or not.\n\n#Definition#:\n- Drivelology: Statements that appear logically coherent but contain deeper, often paradoxical meanings.\nThese challenge conventional interpretation ... [TRUNCATED] ... ology.\n\n#Output Format#:\nYou should try your best to answer \"Yes\" if the given input text is Drivelology, otherwise specify \"No\".\nThe answer you give MUST be \"Yes\" or \"No\"\".\n\n#Input Text#: A: Name? B: Henry. A: Age? B: E-N-R-Y.\n#Your Answer#:"
        }
      ]
    }
  ],
  "target": "YES",
  "id": 0,
  "group_id": 0,
  "metadata": {
    "answer": "YES"
  }
}

Note: Some content was truncated for display.

Prompt Template#

Prompt Template:

{question}
Few-shot Template 
{question}

Usage#

Using CLI#

evalscope eval \
    --model YOUR_MODEL \
    --api-url OPENAI_API_COMPAT_URL \
    --api-key EMPTY_TOKEN \
    --datasets drivel_binary \
    --limit 10  # Remove this line for formal evaluation

Using Python#

from evalscope import run_task
from evalscope.config import TaskConfig

task_cfg = TaskConfig(
    model='YOUR_MODEL',
    api_url='OPENAI_API_COMPAT_URL',
    api_key='EMPTY_TOKEN',
    datasets=['drivel_binary'],
    limit=10,  # Remove this line for formal evaluation
)

run_task(task_cfg=task_cfg)