Cube-Space AutoEncoder

Closed weights MIT-IBM Watson AI Lab April 2020

No estimate

No hardware requirements for this model

The weights for this model have not been published, so it cannot be downloaded or run on your own hardware at any size. It is reachable only through its provider, and no graphics card changes that.

On record

Full specification

Everything on record for this model. Most of it describes how it was trained rather than how it runs — useful context for judging how much work went into it, and how it compares with models built at a different scale.

Origin

Who built this model, where, and when it was published.

Organisation
MIT-IBM Watson AI Lab
Organisation type
Academia,Industry
Country
United States of America
Published
27 April 2020
Authors
Masataro Asai, Christian Muise

What it does

The problem areas the model was built for. A model can carry several of each.

Domain
Vision, Search
Task
Visual puzzles

Size

How large the model is and how much data it was trained on. Parameters are the figure that decides whether it fits on a given graphics card.

Training data
4,243,600,000 tokens

"Finally, we tested Mandrill 15-puzzle, a significantly more challenging 4x4 variant of the sliding tile puzzle (Figure 1). We trained the network with more hyperparameter tuning it- erations (300) and a larger training set (50000). We gener- ated l = 14, 21 instances (20 each) and ran the system (Ta- ble 2, bottom right). "

Training compute

The arithmetic performed to train the model, measured in floating-point operations. It is a measure of what the training run cost, not of how fast the finished model answers you.

Training compute
1.1 × 10¹⁷ FLOP

(1) * (4113 * 10**9) * (24 * 3600) * (0.3) = 106608960000000000 (num gpu) * (peak flops) * (time in seconds) * (assumed utilization rate) = "For each domain, we searched for 100 iterations (≈15min/iter, 24 hours total) on a Tesla K80" 4.113 TFLOPS from https://www.techpowerup.com/gpu-specs/tesla-k80.c2616

How it was established
Hardware

The training run

What it physically took to train: which chips, how many, for how long, and what that drew from the wall.

Training hardware
NVIDIA Tesla K80
Chips used
1
Chip-hours
24
Wall-clock time
24 hours

"For each domain, we searched for 100 iterations (≈15min/iter, 24 hours total) on a Tesla K80"

Power draw
337 W

How it is classified

Labels the source dataset applies when tracking notable models, and how confident it is in the entry.

Record confidence
Confident
Citations
56

Sources

Where this record came from and when it was last checked.

Reference
Learning Neural-Symbolic Descriptive Planning Models via Cube-Space Priors: The Voyage Home (to STRIPS)
Last updated
28 November 2025

What the numbers mean

Background

Cube-Space AutoEncoder was published by MIT-IBM Watson AI Lab, in United States of America, in April 2020. The organisation is categorised as academia,Industry.

It works in Vision, Search, and is recorded as doing visual puzzles.

This is a closed model: the trained values stayed with whoever produced them, and there is no local version to run.

Training and provenance

Training it took roughly 1.1 × 10¹⁷ FLOP of computation, on NVIDIA Tesla K80 — a measure of what producing the model cost, not of how fast it answers.

It was trained on about 4,243,600,000 tokens of text.

Answers

Cube-Space AutoEncoder — common questions

01

How many parameters does Cube-Space AutoEncoder have?

No parameter count has been published for Cube-Space AutoEncoder, which is why no memory or speed figure appears on this page.

02

Who created Cube-Space AutoEncoder?

Cube-Space AutoEncoder was published by MIT-IBM Watson AI Lab, based in United States of America, categorised as academia,Industry.

03

When was Cube-Space AutoEncoder released?

Cube-Space AutoEncoder was published in April 2020. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

04

What is Cube-Space AutoEncoder used for?

Cube-Space AutoEncoder works in Vision, Search, and is recorded as handling visual puzzles. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

05

How much compute was used to train Cube-Space AutoEncoder?

Around 1.1 × 10¹⁷ FLOP, on NVIDIA Tesla K80. That measures what producing the model cost and says nothing about how quickly it answers once trained — inference speed comes from memory bandwidth, not from the training budget.

06

What GPU do I need to run Cube-Space AutoEncoder?

None. Cube-Space AutoEncoder is a closed model — its weights were never published, so it cannot be downloaded or run on your own hardware at any price. It is reachable only through its provider.

07

Is Cube-Space AutoEncoder open source?

The licensing for Cube-Space AutoEncoder was never recorded in our source data. We treat unstated licensing as closed, because an unrecorded licence is not one to rely on.

Source

Original publication

Record last updated 28 November 2025

The other direction

Looking at it from the other side?

This page starts from the model. If you already own a card and want to know everything it will run, start from the hardware instead.