Quantum
Agentic MLIR-to-QIR Translation with Dual-Backend Testing: A Hybrid Deterministic and LLM Approach
Publication: quant-ph updates on arXiv.org
Author: Sharmin Afrose, Vicente Leyton-Ortega, Narasinga Rao Miniskar, Elaine Wong, Travis S. Humble
Source published: 20/11/2026, 05:00:00 UTC
Briefing created: 20/11/2026, 05:38:14 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 20/11/2026, 05:13:25 UTC.
Briefings
A hybrid MLIR-to-QIR translation system combines deterministic compilation for known dialects with LLM-based agentic translation for unknown ones. The pipeline uses dual-backend testing via total variation distance to validate outputs. Results show 95% TVD similarity on 70 circuits across three dialects, with four of six models successfully generating valid QIR on an unseen fault-tolerant dialect.
Key points
- The system selects deterministic compilation for known grammars like Catalyst and Quake, while using LLM agentic paths for unseen dialects.
- A dual-backend pipeline compares generated QIR against source MLIR using total variation distance to ensure validity.
Context
Quantum frameworks define their own MLIR dialects, requiring hand-written passes for porting circuits to QIR runtimes. This research provides a hybrid approach to automate translation across multiple dialects.
Limitations
Coverage is limited to 41 circuits across three dialects and 1-100 qubits. Results are based on a preprint (arXiv:2611.25106v1) and include uncertainty regarding the first iteration of the unseen fault-tolerant dialect model.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 20/11/2026, 05:38:14 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeif442l4rvxczycyknqzycxpflsbbf33ehguxbh4geipiod5g5zivm