Ardea/Icarus-dataset

Icarus

A unified multi-modal curriculum dataset for evolutionary neural architecture search. Every row is one self-contained Task = {meta, support, query}, where support and query are lists of (input_Field, output_Field) pairs. The inner loop trains on support; fitness is scored on query. Support is non-empty for every task. Encoders read the Field descriptor (axes, value_type, n_classes, value_range, mask); mask is True where a value is padding/ignored. meta.class_names, when present, names the target classes (e.g. fsd50k sound events); it is viewer/eval metadata only, never read by the loader or encoder, so the dataset stays structural.

Built by the generator repo https://github.com/ArdeaAI/Icarus-Dataset, which also holds the explorer (uv run explore), the build pipeline, and the full reference tooling. This card and the one self-contained file shipped beside the data (icarus.py: the loader plus a reference encoder, no dependencies on this repo) are everything you need.

Rungs (18-rung difficulty ladder; one config rung_<N> each)

split is how support/query is formed: native rungs ship a built-in split (ARC train/test, XOR's degenerate support==query) and the loader keeps it as-is; bucketed rungs are pooled and the loader carves them at load time via support_fraction (so the same data streams at any ratio).

rung	name	split	support_fraction	bucket_size	window	config
1	XOR	native	-	native	-	rung_1
2	Parity-N	bucketed	0.8	64	-	rung_2
3	Two-spirals	bucketed	0.8	100	-	rung_3
4	Pole (Markov)	bucketed	0.8	100	8	rung_4
5	Double-pole (no velocity)	bucketed	0.8	100	8	rung_5
6	MNIST / Fashion-MNIST	bucketed	0.8	100	-	rung_6
7	CIFAR-10 / CIFAR-100	bucketed	0.8	100	-	rung_7
8	NB360 ecg	bucketed	0.8	64	-	rung_8
9	NB360 satellite	bucketed	0.8	64	-	rung_9
10	NB360 ninapro	bucketed	0.8	64	-	rung_10
11	NB360 spherical	bucketed	0.8	64	-	rung_11
12	NB360 cosmic	bucketed	0.8	64	-	rung_12
13	NB360 darcy_flow	bucketed	0.8	64	-	rung_13
14	NB360 psicov	bucketed	0.8	64	-	rung_14
15	NB360 fsd50k	bucketed	0.8	64	-	rung_15
16	NB360 deepsea	bucketed	0.8	64	-	rung_16
17	RAVEN / PGM	bucketed	0.8	64	-	rung_17
18	ARC-AGI v1	native	-	native	-	rung_18

Explorer

Usage

Stream rows (the datasets library only)

Each row is a serialized task. icarus.py (shipped beside the data) reconstructs a whole Task:

from datasets import load_dataset
from icarus import deserialize_task  # the single file shipped with this dataset

stream = load_dataset("Ardea/Icarus-dataset", name="rung_6", streaming=True, split="train")
task = deserialize_task(next(iter(stream)))  # -> Task(meta, support, query)

Vendored loader (whole Task objects, MAXES selection)

from icarus import IcarusDataset

# n_samples is the TOTAL examples per task; support_fraction (bucketed rungs only) sets the support share.
dataset = IcarusDataset(rungs=(3, 6, 18), n_tasks=100, n_samples=50, support_fraction=0.8, hf_repo="Ardea/Icarus_dataset")
task = dataset[0]  # a whole Task split into support/query at load time

Reference encoder (structural Task -> tensors)

The encoder is a swappable reference living in the same icarus.py; a real consuming model brings its own featurization and loss.

from icarus import IcarusDataset, Level0Encoder, encode_task

task = IcarusDataset(rungs=(6,), n_tasks=1, n_samples=20, hf_repo="Ardea/Icarus_dataset")[0]
encoded = encode_task(task, Level0Encoder(max_flat_dim=4096))
# encoded.support_input -> (tensor, descriptor); encoded.support_target -> (tensor, mask, descriptor)

Images are stored as uint8 with value_range (0, 255); the reference encoder normalizes.

Sources, attribution & licensing

Ardea-authored material (the build pipeline, the vendored icarus.py, the schema, and the generated rungs 1-5) is released under the MIT License with Attribution (see LICENSE.md): redistribution must retain the copyright notice and state "This software is based on work originally developed by Ardea AI Corp."

Rungs 6-18 redistribute third-party data, and each source keeps its own upstream license (listed below and in SOURCES.md). You must comply with the upstream license of any rung you use or redistribute. Where a license shows as unknown or - it has not been verified for redistribution; confirm the upstream terms before relying on that rung.

rung	family	source	revision	license	acquisition
1	xor	generated	-	n/a	generated
2	parity	generated	-	n/a	generated
3	two_spirals	generated	-	n/a	generated
4	pole	generated	-	n/a	generated
5	double_pole_no_velocity	generated	-	n/a	generated
6	mnist	ylecun/mnist	-	mit	hf-cache
6	fashion_mnist	zalando-datasets/fashion_mnist	-	mit	hf-cache
7	cifar10	uoft-cs/cifar10	-	unknown	hf-cache
7	cifar100	uoft-cs/cifar100	-	unknown	hf-cache
8	ecg	/Volumes/Pickles/NAS_Bench_360/NAS-Bench-360/ecg	-	see NAS-Bench-360 (rtu715/NAS-Bench-360)	local
9	satellite	/Volumes/Pickles/NAS_Bench_360/NAS-Bench-360/satellite	-	see NAS-Bench-360 (rtu715/NAS-Bench-360)	local
10	ninapro	/Volumes/Pickles/NAS_Bench_360/NAS-Bench-360/ninapro	-	see NAS-Bench-360 (rtu715/NAS-Bench-360)	local
11	spherical	/Volumes/Pickles/NAS_Bench_360/NAS-Bench-360/spherical	-	see NAS-Bench-360 (rtu715/NAS-Bench-360)	local
12	cosmic	/Volumes/Pickles/NAS_Bench_360/NAS-Bench-360/cosmic	-	see NAS-Bench-360 (rtu715/NAS-Bench-360)	local
13	darcy_flow	/Volumes/Pickles/NAS_Bench_360/NAS-Bench-360/darcyflow	-	see NAS-Bench-360 (rtu715/NAS-Bench-360)	local
14	psicov	/Volumes/Pickles/NAS_Bench_360/NAS-Bench-360/psicov/protein	-	see NAS-Bench-360 (rtu715/NAS-Bench-360)	local
15	fsd50k	/Volumes/Pickles/NAS_Bench_360/NAS-Bench-360/audio	-	see NAS-Bench-360 (rtu715/NAS-Bench-360)	local
16	deepsea	/Volumes/Pickles/NAS_Bench_360/NAS-Bench-360/deepsea	-	see NAS-Bench-360 (rtu715/NAS-Bench-360)	local
17	raven	HuggingFaceM4/RAVEN	-	unknown	hf-cache
17	pgm	HuggingFaceM4/PGM	-	unknown	hf-cache
18	arc	Ardea/arc_agi_v1	-	apache-2.0	hf-cache