RaindropA practical guide to choosing decision modelsGitHub ↗

Similar distractors · Add fewer similar but irrelevant items variant

Original question · Other versions of this question · All variants

This version keeps the original reference answer. Its context, question instructions or answer choices were changed as described below.

Changed fields: context.

How the edits were constructed

Context

“A person chose curtains for a living room.”

Remy was decorating their room and bought new sheets.

Question

How would you describe Remy?

Answer choices

  • A. As someone that likes to decorate
  • B. proud
  • C. As someone who's asked about their room

Reference answer: A.

Exact evaluated input as JSON
{
  "state": "“A person chose curtains for a living room.”\n\nRemy was decorating their room and bought new sheets.",
  "question": "How would you describe Remy?",
  "choices": {
    "A": "As someone that likes to decorate",
    "B": "proud",
    "C": "As someone who's asked about their room"
  }
}

Recorded model answers

The original column uses the control response paired with this variant. Missing responses and inputs outside a model’s limit remain visible.

Model Original answer Variant answer Change
GLiDE v1.2 A — As someone that likes to decorate A — As someone that likes to decorate Same answer
Decider v1.1 A — As someone that likes to decorate A — As someone that likes to decorate Same answer
Jev A — As someone that likes to decorate A — As someone that likes to decorate Same answer
Microsoft Decision-1 A — As someone that likes to decorate A — As someone that likes to decorate Same answer
Clef A — As someone that likes to decorate A — As someone that likes to decorate Same answer
Luna Decisions A — As someone that likes to decorate A — As someone that likes to decorate Same answer
Clef Flash A — As someone that likes to decorate A — As someone that likes to decorate Same answer
Strands Decider A — As someone that likes to decorate A — As someone that likes to decorate Same answer
GLiNER2.5 Decide A — As someone that likes to decorate A — As someone that likes to decorate Same answer
Laya A — As someone that likes to decorate A — As someone that likes to decorate Same answer
CalDec v1 A — As someone that likes to decorate A — As someone that likes to decorate Same answer
Identifiers and recorded conditions

Variant ID: socialiqa/663709fec0fa8c9ab71a/similar_distractors/low/clean-source_first_social_quoted_companions_full_0207_v2

Input fingerprint: f15fbb7e80eeef1db94cb834b60c629b5f46836350e27dc62a7710b7d8b98b9e

Change Recorded version
Similar distractors Add fewer similar but irrelevant items

Source attribution and dataset terms are linked from the original question.

Type to search questions, variants and documentation.