MarkBench-Final: Deterministic Graph-Topology & State Consistency Benchmark
📌 Overview
MarkBench-Challenger is a rigorous evaluation suite designed to test advanced frontier LLMs on multi-constraint satisfaction, structural state tracking, and deterministic algorithmic execution.
Unlike traditional benchmarks that evaluate verbal reasoning or isolated coding tasks, MarkBench challenges an agent's capability to maintain topological precision and mathematical… See the full description on the dataset page: https://huggingface.co/datasets/Axiom-AI/MarkBench-Challenger.