Temporal order

What differs between trials?

temporal_order presents A then B or B then A. The first order requires left; the second requires right. Both then present the same terminal C before go. The four receptors are A, B, C and go; two effectors vote during response. There is one neural frame per world tick.

first item (8) → gap (8) → second item (8) → terminal C (8) → response (8)

from brainlesslab import TaskSpec
from brainlesslab.tasks import definition, resolve

task = TaskSpec("temporal_order", options={"gap": 8})
print(resolve(task))
print(definition(task))

Expect four inputs, two effectors and horizon 40. Three presets use gaps 0, 8 and 32.

Which shortcuts remain?

The total counts of A and B are equal in either order. A bag-of-items reference therefore cannot distinguish the answer by counts; a last-item reference sees the same terminal C. An observation-based oracle retains the first item.

Retaining the first item is sufficient for this task. Success does not require a general representation of sequence order or a grammar. input_mode="hidden" removes A and B but leaves terminal C and go visible, preserving the shared ending and schedule.

probe_accuracy is one binary result per full trial, with left selected on tied effector totals. Analytic chance is 0.5 and warm-up is zero. Report the gap and input visibility, then compare first-item, bag-of-items and last-item references as declared conditions.

Continue with recall to isolate retention duration, or decoding to inspect emitted information.