DET

DET (Detective) applies Divide Hierarchical Clustering to the seed list, learns which dimensions exhibit stable values, and expands those regions while maintaining an adaptive scanning frontier. Leaf patterns generate candidates by relaxing low-entropy dimensions first.

• Reference: Placeholder citation (coming soon).

Train

rmap train --seeds seeds/hitlist.txt --output models/det.bin det \
  --m 4 --min-leaf-size 10

Generate

rmap generate --model models/det.bin --count 200000 \
  --unique --output det.txt

Configuration

• --m <usize> – bits per dimension when projecting addresses into vectors (default 4, producing 32 dimensions).
• --min-leaf-size <usize> – minimum number of seeds allowed in a leaf before it splits (default 10).
• --seed <u64> – optional RNG seed (default 0).

Model notes

• Smaller --m values increase dimensionality and may uncover finer-grained prefixes at the cost of runtime.
• The serialized model captures both the space tree and per-leaf decision structures (DS/TS) so deterministic replay is possible across tool versions.