[challenge]: Faster Expert TensorCircuit-NG Solutions for ORBIT-Q benchmark

Released by

Shi-Xin Zhang, IOP-CAS

Contact email

shixinzhang@iphy.ac.cn

Method

Quantum Circuit Simulation

Challenge issue

ORBIT-Q contains expert TensorCircuit-NG reference solutions for 12 quantum-programming tasks. These solutions are semantically correct and serve as reference artifacts, but they are not necessarily runtime-optimal. This challenge asks participants to improve the human expert TensorCircuit-NG solutions themselves or together with AI agents, with the goal of reducing the evaluator-reported end-to-end runtime while preserving the original task semantics and verifier correctness.

The target is to produce faster expert-level TensorCircuit implementations for the same ORBIT-Q problem definitions. A valid improvement must keep the original run_solution(config) contract, pass the original functional checks, respect the required quantum-computing semantics, and avoid shortcutting the task through hardcoded answers, evaluator-specific hacks, or replacing the intended TensorCircuit-NG computation with an unrelated raw simulator. If necessary, the improvement can go with the PR on TensorCircuit-NG framework itself.

A successful submission should report runtime against the current ORBIT-Q expert reference solution on the same evaluator and hardware/software environment (averaged over multiple runs).

Why this may lead to research output: it can produce a stronger expert-performance baseline for ORBIT-Q and clarify the true efficiency frontier of framework-native quantum-programming solutions. This is important because agent performance should be compared against high-quality human expert artifacts, not merely the first correct reference implementation.

参考 ORBIT-Q benchmark 的公众号介绍

Start by reviewing the ORBIT-Q benchmark and its 12 TensorCircuit-NG expert reference solutions, then identify how the evaluator measures end-to-end runtime. Any result must preserve the run_solution(config) contract, pass the original functional checks and quantum semantics, and report averaged runtime against the current reference on the same environment.