RaindropA practical guide to choosing decision modelsGitHub ↗

Learning and induction · question 25

Back to Learning and induction · All capabilities

Question ID: list_functions/af2a611ae53f2fe68bf8
Source: List Functions
Input format: adapted choices

Explore this question’s altered versions

Context

[{"input": "[0]", "output": "[]"}, {"input": "[1, 6]", "output": "[6]"}, {"input": "[7]", "output": "[]"}, {"input": "[0, 8]", "output": "[8]"}, {"input": "[9, 7, 1, 8, 2, 3]", "output": "[7, 1, 8]"}, {"input": "[1, 2, 3]", "output": "[2, 3]"}, {"input": "[4]", "output": "[]"}, {"input": "[6, 8]", "output": "[8]"}]

Question

Apply the same inferred rule to this new input: [9, 7, 1, 5, 2, 0, 6, 4]

Answer choices

  • A: [9, 7, 1]
  • B: [7, 1, 5]
  • C: [1, 2, 3]
  • D: [10, 8, 2]

Reference answer: [7, 1, 5] (B)

Exact evaluated input as JSON
{
  "state": "[{\"input\": \"[0]\", \"output\": \"[]\"}, {\"input\": \"[1, 6]\", \"output\": \"[6]\"}, {\"input\": \"[7]\", \"output\": \"[]\"}, {\"input\": \"[0, 8]\", \"output\": \"[8]\"}, {\"input\": \"[9, 7, 1, 8, 2, 3]\", \"output\": \"[7, 1, 8]\"}, {\"input\": \"[1, 2, 3]\", \"output\": \"[2, 3]\"}, {\"input\": \"[4]\", \"output\": \"[]\"}, {\"input\": \"[6, 8]\", \"output\": \"[8]\"}]",
  "question": "Apply the same inferred rule to this new input: [9, 7, 1, 5, 2, 0, 6, 4]",
  "choices": {
    "A": "[9, 7, 1]",
    "B": "[7, 1, 5]",
    "C": "[1, 2, 3]",
    "D": "[10, 8, 2]"
  }
}

Recorded model answers

Model Answer Outcome
Astra [7, 1, 5] Correct
Opus 5.5 [7, 1, 5] Correct
GLiDE v1.2 [7, 1, 5] Correct
Decider v1.1 [7, 1, 5] Correct
Jev [7, 1, 5] Correct
Microsoft Decision-1 [7, 1, 5] Correct
Clef [7, 1, 5] Correct
Luna Decisions [7, 1, 5] Correct
Clef Flash [9, 7, 1] Incorrect
Strands Decider [7, 1, 5] Correct
GLiNER2.5 Decide [9, 7, 1] Incorrect
Laya [9, 7, 1] Incorrect
CalDec v1 [9, 7, 1] Incorrect

Source and sharing

Licensed upstream task: c011. All eight supplied examples and the evaluated test input/output match the original BIG-bench data.

Attribution: Joshua S. Rule, Steven T. Piantadosi, Joshua B. Tenenbaum; BIG-bench contributors; Linlu Qiu et al. (selection).
Terms: Apache-2.0. Source sharing decision.
The capability assignment and choice-format normalization are part of this experiment. Source data retain their own terms.

Exact source locator and input fingerprint
{
  "source_locator": {
    "path": "data/list_function.jsonl",
    "record": "10/0",
    "kind": "source_file_record"
  },
  "source_revision": "0eae21f3d3014b92be34c80b23c9120a7b15a51f",
  "input_hash": "952d62bf795fc644ecbdc5e211d7b06b934875da41f43ad48c2ff61d919d9251"
}

Type to search questions, variants and documentation.