Dataset Viewer
Auto-converted to Parquet Duplicate
benchmark_name
string
description
string
tasks
list
Relational Language Reasoning Benchmark
Tasks designed to test whether relational language availability affects reasoning performance. Each task has a version that forbids relational words and a normal version. All tasks have exactly one correct ordering.
[ { "id": 1, "prompt_no_relational": "Solve this puzzle but you are ABSOLUTELY NOT allowed to use any relational words such as: next to, left, right, order, between, before, after, same, different.\n\nPuzzle: Four colors are in a line: red, blue, green, yellow.\nClue 1: Blue is immediately left of Red.\nClue ...

No dataset card yet

Downloads last month
9