Three LECS papers head to IEEE NEWCAS 2026

The lab is bringing three papers to the IEEE Interregional NEWCAS (New Circuits and Systems) Conference 2026, held in Chicoutimi, Québec, Canada on June 21–24, 2026. Two of the works push on quantum computing — one at the system-interconnect level, one at the state-preparation level — while the third tackles efficiency in ReRAM-based graph accelerators. Here is a short tour of each.

Synthesizing the interconnect for distributed quantum computers

Milad Eslaminia presents Synthesis of Application-Specific Quantum Network-on-Chip. Distributed quantum computing scales by wiring multiple QPUs together, but the payoff is bounded by the inter-QPU communication budget: communication qubits, entanglement generation, and link capacity. The paper makes the case for tailoring the interconnect to the workload — an application-specific Quantum NoC (QNoC).

The work introduces an automated QNoC synthesis flow for multi-QPU architectures built on dedicated communication qubits with LTM interfaces and a silicon-photonic entanglement distribution network. It explores candidate subnet mappings and link configurations, evaluates each through transpilation, and ranks them with a weighted score that balances communication usage, SWAP overhead, and execution time. Across benchmark circuits, the synthesis surfaces workload-dependent trade-offs and the kind of architectural insight that a one-size-fits-all interconnect would hide.

Preparing GHZ states that survive the noise

Jean-Baptiste Waring presents Robust GHZ State Preparation via Majority-Voted Boundary Measurements. High-fidelity Greenberger–Horne–Zeilinger (GHZ) states are hard to prepare on real hardware: gate errors and decoherence accumulate, and measurement errors propagate through classical feedforward in dynamic circuits.

The paper introduces Group-Majority-Voting (Group-MV), a dynamic-circuit protocol that partitions an arbitrary coupling graph, prepares local GHZ states in parallel, and fuses them through majority-voted mid-circuit measurements. Voting over redundant boundary links suppresses the measurement errors that would otherwise corrupt the fusion. Evaluated on simulated Heavy-hex and Grid topologies from 30 to 60 qubits under a realistic noise model, Group-MV generalizes to arbitrary GHZ sizes on arbitrary topologies, reaching 2.4× higher fidelity than the Line Dynamic method while staying within 3% of the noiseless unitary baseline.

Reusing patterns in ReRAM graph accelerators

Masoud Rahimi presents Graph Canonization for Efficient Pattern Reuse in ReRAM-Based Graph Accelerators. In-situ computation in ReRAM crossbars is a strong fit for large, sparse graphs, but pattern-based designs hit a wall: the number of unique subgraph patterns explodes as the crossbar grows, starving the static engines and dragging down throughput.

The paper restores pattern reuse with two ideas. A degree-aware reordering algorithm cuts the number of subgraphs that large crossbars have to handle, and a canonical pattern normalization algorithm rewrites subgraphs into canonical forms so that far fewer distinct patterns remain. On representative real-world graphs the combination delivers up to 9.4× speedup over a state-of-the-art pattern-based ReRAM accelerator.

See the papers

• Synthesis of Application-Specific Quantum Network-on-Chip — Milad Eslaminia, Sébastien Le Beux
• Robust GHZ State Preparation via Majority-Voted Boundary Measurements — Jean-Baptiste Waring, Sébastien Le Beux, Christophe Péré
• Graph Canonization for Efficient Pattern Reuse in ReRAM-Based Graph Accelerators — Masoud Rahimi, Sébastien Le Beux

Congratulations to the authors — see you in Chicoutimi.

Comments & corrections

If you have questions about this note, or you spot something we got wrong, please write to the author directly. We post addenda to articles when a correction is warranted.