II.
Benchmark overview
Reference · livebenchmark:arc-challenge
ARC-Challenge overview
The AI2 Reasoning Challenge (ARC) — Challenge set — is a collection of 2,590 grade-school-level multiple-choice science questions selected to be unanswerable by simple retrieval or word-correlation baselines, introduced by Clark et al. (Allen AI, 2018).
Attributes
displayName
ARC-Challenge
homepageUrl
kind
model-only
targetsKind
ModelVersion
description
The AI2 Reasoning Challenge (ARC) — Challenge set — is a collection
of 2,590 grade-school-level multiple-choice science questions
selected to be unanswerable by simple retrieval or word-correlation
baselines, introduced by Clark et al. (Allen AI, 2018).
Outgoing edges
None.
Incoming edges
for_benchmark1
scored_against1