Lookup tables

A lookup-table component maps a finite encoded domain to an encoded range. Its contract records signed interpretation, table cardinality, invalid indices, material ownership, one-time use, communication, and whether selection is data-oblivious under the stated threat model.

Tables are useful for nonlinearities only when domain bounds and total storage remain explicit. Approximation and model-quality evidence belong to the numeric policy, not the table name.

BinaryTableGatedMultiplyQ7 exposes the bounded Q7 gated-multiply method as public configuration; it does not expose generic user-defined tables or evaluator labels. See the exact nonlinear API and research evidence boundary.

API: Nonlinear objects and signatures

Python SDK example

from pllm.nonlinear import BinaryTableGatedMultiplyQ7

method = BinaryTableGatedMultiplyQ7()
assert method.to_spec() == {
    "component": "pllm/binary-table/v1",
    "params": {},
}

The object selects a compiler component. It does not evaluate a table or establish that a complete model can use the method.