Neural Clifford Synthesis
Synthesise 6 qubit Clifford Quantum Circuits Online using a Reinforcement Learning Agent. Edit the tableau in QASM mode and compare against baselines. Running the agent in rollout mode often finds an optimal circuit.
Paper
Bravyi et al Benchmark
Target Tableau
Ready
Customize
Seed
Walk
Random walk
Identity
Encoding
Apply
Compact
↻
RL Solution
Gates
--
1Q
--
2Q
--
No run
Rel16 new - 64 rollouts, inverse
Decoder
64 rollouts
RL policy
Paper model (6FC latest)
New model (relgate16)
Relgate16 top-up 2B
Reltransformer gate-cost latest
7FC latest
10FC latest
Mode
Greedy
Rollout
Max steps
inverse
Rollouts
Start T
End T
Schedule
Linear
Fixed
Cosine
Exponential
Solve
Stop
0/0