Variable-length Gray codes for the Natural Numbers
arXiv:2607.16088
2026
Architecture
2 ideas extracted · analyzed Aug 30, 2026
What the math gives to ML
The paper gives a constructive bijection from natural numbers to all finite digit strings whose length grows monotonically and whose consecutive integers differ by exactly one edit. The transferable asset is not compression alone, but the combination of self-delimiting variable length, complete coverage including leading-zero strings, and local transitions at integer increments. A practical ML use is to represent counters and numeric categorical values as digit sequences under this code, then exploit the known one-edit transition structure in embeddings, auxiliary losses, or incremental inference rather than treating each integer as an unrelated token.
Ideas from this paper
Unverified
2026
Replace a learned embedding for each integer with a compositional embedding of its variable-length Gray codeword. Add an auxiliary constraint that numerically adjacent values have nearby representations, while preserving the ordinary task loss so that the model can learn when numerical adjacency matters.
Useful6/10
Difficulty4/10
Novelty8/10
Unverified
2026
Use the one-edit Hamiltonian walk as an explicit state machine for counters in autoregressive models or world models. Instead of regenerating and re-embedding an entire numeric string after every increment, update only the digit that changes, or append the single leading digit at a block transition.
Useful5/10
Difficulty6/10
Novelty9/10