ARC-C

The 'ARC-C' dataset/benchmark contains a collection of reasoning tasks designed to evaluate the generalization capabilities of models, particularly in the context of multi-step reasoning.

Papers using ARC-C (2)

Punching Above Their Weight: Classification-Head Fine-Tuning of Tiny Language Models (TLMs) for Verifiable Multiple-Choice Tasks2026CPL: Critical Plan Step Learning Boosts LLM Generalization in Reasoning Tasks2024