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: question instructions.

How the edits were constructed

Context

Speaker_1: He was very sad. Speaker_2: Yeah, He was sad like a widow.

Question

Choose the interpretation or answer requested by the text.

«What does “perfectly imperfect” mean?»

Answer choices

  • A. Agrees
  • B. Sarcastic

Reference answer: A.

Exact evaluated input as JSON
{
  "state": "Speaker_1: He was very sad.\nSpeaker_2: Yeah, He was sad like a widow.",
  "question": "Choose the interpretation or answer requested by the text.\n\n«What does “perfectly imperfect” mean?»",
  "choices": {
    "A": "Agrees",
    "B": "Sarcastic"
  }
}

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 — Agrees B — Sarcastic Different answer
Decider v1.1 A — Agrees A — Agrees Same answer
Jev A — Agrees B — Sarcastic Different answer
Microsoft Decision-1 A — Agrees B — Sarcastic Different answer
Clef A — Agrees B — Sarcastic Different answer
Luna Decisions A — Agrees A — Agrees Same answer
Clef Flash A — Agrees A — Agrees Same answer
Strands Decider A — Agrees B — Sarcastic Different answer
GLiNER2.5 Decide B — Sarcastic B — Sarcastic Same answer
Laya B — Sarcastic B — Sarcastic Same answer
CalDec v1 A — Agrees A — Agrees Same answer
Identifiers and recorded conditions

Variant ID: pub/d56e739cd4f847b2da94/similar_distractors/low/clean-source_first_referent_free_near_full_0285

Input fingerprint: 0d2ced17922e110d1b11cf416aa0b02a94c2cff116879468fb10476d34d0ab3d

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.