RaindropA practical guide to choosing decision modelsGitHub ↗

Evidence position · Move the allegation to the middle 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

So Boris has let 150,000 people die, made a hash of Brexit and lied through his teeth but still thinks he deserves a boozy bash! Nothing surprises me about him any more sadly.

Question

Is this message intended to be sarcastic?

Answer choices

  • yes. Sarcastic
  • no. Not sarcastic

Reference answer: no.

Exact evaluated input as JSON
{
  "state": "So Boris has let 150,000 people die, made a hash of Brexit and lied through his teeth but still thinks he deserves a boozy bash! Nothing surprises me about him any more sadly.",
  "question": "Is this message intended to be sarcastic?",
  "choices": {
    "yes": "Sarcastic",
    "no": "Not 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 no — Not sarcastic no — Not sarcastic Same answer
Decider v1.1 no — Not sarcastic no — Not sarcastic Same answer
Jev no — Not sarcastic no — Not sarcastic Same answer
Microsoft Decision-1 no — Not sarcastic no — Not sarcastic Same answer
Clef no — Not sarcastic no — Not sarcastic Same answer
Luna Decisions yes — Sarcastic yes — Sarcastic Same answer
Clef Flash no — Not sarcastic no — Not sarcastic Same answer
Strands Decider yes — Sarcastic yes — Sarcastic Same answer
GLiNER2.5 Decide yes — Sarcastic yes — Sarcastic Same answer
Laya yes — Sarcastic no — Not sarcastic Different answer
CalDec v1 yes — Sarcastic yes — Sarcastic Same answer

Ordering comparison: literal local or custom unit permutation. Only the exact matched comparison sets contribute to the reported ordering contrasts.

Identifiers and recorded conditions

Variant ID: isarcasmeval/3abe6ec5adac85690586/evidence_position/native_allegation_middle/clean-source_first_native_unit_roles_strong_shuffle_full_0380

Input fingerprint: 320c2f0a7e766aea5592859aab3f7abe8be92479fb71a7271709c34aae371595

Change Recorded version
Evidence position Move the allegation to the middle
Evidence order Reorder alternating parts of the allegation

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

Type to search questions, variants and documentation.