Benchmarks

Every number, losses included

Each figure below is generated from a recorded run and lives in paper/neurips/generated/numbers.tex. Nothing here is hand-typed.

Continual stream, 100 sequential tasks

Immediate accuracy is measured at the moment of learning; retained accuracy after the full stream. Chance is 0.0625.

SystemImmediateRetainedWhat it isolates
Page memory (Hebb)1.0001.000Allocation and a write
Episodic store0.9880.988Allocation, no write
Fast weights0.6670.072A write, no allocation

Full fine-tuning and LoRA sit with fast weights, for the same reason: one shared state, overwritten.

Allocation predicts retention

Varying only the page count across a 128× range (1–128 pages) leaves immediate accuracy at 1.000 throughout, while retained accuracy moves from 0.068 to 1.000. The fraction of memories still holding a page of their own predicts retained accuracy to within 0.0039 across all 6 settings, with no free parameters.

Addressing — the open problem

Over 198 rules, learned content addressing sits on a frontier rather than clearing it.

MeasureValue
Best false-merge rate0.353
Best missed-revisit rate0.222
Best retrieval@10.717
Key separation, meta-trained0.897
Key separation, untrained0.425

Where we lose

On vision, a plain episodic store beats page memory outright: 0.994 against our 0.697. That is on the home page footer for the same reason it is here.

BenchmarkPage memoryEpisodic
Vision0.6970.994
Compositional0.4580.374

Footprint

Per memory, against an episodic store on the same stream — not against a per-tenant adapter, which is a different comparison.

StoreBytes per memory
Page memory1,164
Episodic56,448
Ratio48×