DiffLogic CA fully converges on Game of Life rules — both soft and hard losses converge to zero on all 512 configurations

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• DiffLogic CA fully converges to 16x16 checkerboard target pattern with both soft and hard losses reaching zerofinding0.847
  Core result of pattern generation experiment demonstrating recurrent circuit learning
• DiffLogic CA with 128-bit cell state and 12 steps successfully learns 20x20 lizard pattern, generalizing to 40x40 gridfinding0.758
  Demonstration of DiffLogic CA on complex non-regular shapes with arbitrary memorization requirements
• DiffLogic CA has strong kinship with CAM-8 architecture where each cell uses DRAM for state variables and SRAM for processingclaim0.755
  Authors' architectural analogy between DiffLogic CA and Toffoli-Margolus CAM-8
• DiffLogic CA learned fault tolerance and self-healing behavior without explicit design for these conditionsfinding0.754
  Key finding on robustness — both permanent and temporary cell deactivation handled gracefully
• DiffLogic CA represents, to best of authors' knowledge, the first exploration of differentiable logic gate networks in a recurrent settingfinding0.750
  Novelty claim about the contribution to the field
• Asynchronously trained DiffLogic CA shows greater robustness to 10x10 pixel damage than synchronously trained version, measured by sum of absolute differencesfinding0.748
  Quantitative comparison of synchronous vs asynchronous training for noise resilience
• DiffLogic CA mirrors how biological systems achieve reliability through networks of imperfect componentsclaim0.746
  Authors' analogy between emergent fault tolerance in DiffLogic CA and biological robustness
• Synchronously trained DiffLogic CA circuit succeeds at asynchronous inference without retrainingfinding0.743
  Unexpected result demonstrating robustness of learned circuits beyond their training regime