SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2110.03331 · ICLR · 2022

CLEVA-Compass: A Continual Learning EValuation Assessment Compass to Promote Research Transparency and Comparability

Kristian Kersting, Martin Mundt, Steven Lang, Quentin Delfosse

arXiv · PDF · Open in the Atlas

Code that ran

8 of 10 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (3 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 3 have no contract verdict; 5 failures of our instrument; 2 no recorded run), measured 2026-11-22; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2110.03331.json · badge: /api/ran/2110.03331.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
ml-research/CLEVA-Compass canonical (classifier label)
classifier basis: classifier_2025
8 ran · 2 no recorded run
FunctionStatusWhere it lives
CompassEntry Ran
ran on a synthesized input · constructed an object; no result was computed
ml-research/CLEVA-Compass/create_compass.py
code served (permissive licence) · get_code("070b06b058ccd060")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned an instance of its own class, CompassEntry: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
from dataclasses import dataclass

@dataclass
class InnerLevel:
    """Inner CLEVA-Compass level with method attributes."""
    multiple_models: int
    federated: int
    online: int
    open_world: int
    multiple_modalities: int
    active_data_query: int
    task_order_discovery: int
    task_agnostic: int
    episodic_memory: int
    generative: int
    uncertainty: int

    def __iter__(self):
        """
        Defines the iteration order. This needs to be the same order as defined in the
        cleva_template.tex file.
        """
        for item in [self.multiple_models, self.federated, self.online, self.open_world, self.multiple_modalities, self.active_data_query, self.task_order_discovery, self.task_agnostic, self.episodic_memory, self.generative, self.uncertainty]:
            yield item

@dataclass
class OuterLevel:
    """Outer CLEVA-Compass level with measurement attributes."""
    compute_time: bool
    mac_operations: bool
    communication: bool
    forgetting: bool
    forward_transfer: bool
    backward_transfer: bool
    openness: bool
    parameters: bool
    memory: bool
    stored_data: bool
    generated_data: bool
    optimization_steps: bool
    per_task_metric: bool
    task_order: bool
    data_per_task: bool

    def __iter__(self):
        """
        Defines the iteration order. This needs to be the same order as defined in the
        cleva_template.tex file.
        """
        for item in [self.parameters, self.compute_time, self.mac_operations, self.communication, self.forgetting, self.forward_transfer, self.backward_transfer, self.openness, self.data_per_task, self.task_order, self.per_task_metric, self.optimization_steps, self.generated_data, self.stored_data, self.memory]:
            yield item

@dataclass
class CompassEntry:
    """Compass entry containing color, label, and attributes."""
    color: str
    label: str
    inner_level: InnerLevel
    outer_level: OuterLevel


torch.manual_seed(0)

inner_level = InnerLevel(
    multiple_models=1,
    federated=0,
    online=1,
    open_world=0,
    multiple_modalities=1,
    active_data_query=0,
    task_order_discovery=1,
    task_agnostic=0,
    episodic_memory=1,
    generative=0,
    uncertainty=1
)

outer_level = OuterLevel(
    compute_time=True,
    mac_operations=False,
    communication=True,
    forgetting=False,
    forward_transfer=True,
    backward_transfer=False,
    openness=True,
    parameters=False,
    memory=True,
    stored_data=False,
    generated_data=True,
    optimization_steps=False,
    per_task_metric=True,
    task_order=False,
    data_per_task=True
)

ARGS = {
    "color": "#FF5733",
    "label": "test_entry",
    "inner_level": inner_level,
    "outer_level": outer_level
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type CompassEntry — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a CompassEntry output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
InnerLevel Ran
ran on a synthesized input · constructed an object; no result was computed
ml-research/CLEVA-Compass/create_compass.py
code served (permissive licence) · get_code("5704f6fbebf03b56")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned an instance of its own class, InnerLevel: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
from dataclasses import dataclass

# Set seed for reproducibility
torch.manual_seed(0)

@dataclass
class InnerLevel:
    """Inner CLEVA-Compass level with method attributes."""
    multiple_models: int
    federated: int
    online: int
    open_world: int
    multiple_modalities: int
    active_data_query: int
    task_order_discovery: int
    task_agnostic: int
    episodic_memory: int
    generative: int
    uncertainty: int

    def __iter__(self):
        """
        Defines the iteration order. This needs to be the same order as defined in the
        cleva_template.tex file.
        """
        for item in [self.multiple_models, self.federated, self.online, self.open_world, self.multiple_modalities, self.active_data_query, self.task_order_discovery, self.task_agnostic, self.episodic_memory, self.generative, self.uncertainty]:
            yield item

# Create arguments for InnerLevel
ARGS = {
    "multiple_models": 1,
    "federated": 0,
    "online": 1,
    "open_world": 0,
    "multiple_modalities": 1,
    "active_data_query": 0,
    "task_order_discovery": 1,
    "task_agnostic": 0,
    "episodic_memory": 1,
    "generative": 0,
    "uncertainty": 1,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type InnerLevel — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a InnerLevel output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
OuterLevel Ran
ran on a synthesized input · constructed an object; no result was computed
ml-research/CLEVA-Compass/create_compass.py
code served (permissive licence) · get_code("ba68a5d5e9e187b3")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned an instance of its own class, OuterLevel: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np
from dataclasses import dataclass

torch.manual_seed(0)

ARGS = {
    "compute_time": True,
    "mac_operations": False,
    "communication": True,
    "forgetting": False,
    "forward_transfer": True,
    "backward_transfer": True,
    "openness": False,
    "parameters": True,
    "memory": False,
    "stored_data": True,
    "generated_data": False,
    "optimization_steps": True,
    "per_task_metric": True,
    "task_order": False,
    "data_per_task": True,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type OuterLevel — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a OuterLevel output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
fill_template Ran
ran; our contract draft was wrong, not the code
ml-research/CLEVA-Compass/create_compass.py
code served (permissive licence) · get_code("7893ea67beb14fee")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import math
from dataclasses import dataclass
from typing import List

EV = 15

@dataclass
class InnerLevel:
    """Inner CLEVA-Compass level with method attributes."""
    multiple_models: int
    federated: int
    online: int
    open_world: int
    multiple_modalities: int
    active_data_query: int
    task_order_discovery: int
    task_agnostic: int
    episodic_memory: int
    generative: int
    uncertainty: int

    def __iter__(self):
        """
        Defines the iteration order. This needs to be the same order as defined in the
        cleva_template.tex file.
        """
        for item in [self.multiple_models, self.federated, self.online, self.open_world, self.multiple_modalities, self.active_data_query, self.task_order_discovery, self.task_agnostic, self.episodic_memory, self.generative, self.uncertainty]:
            yield item

@dataclass
class OuterLevel:
    """Outer CLEVA-Compass level with measurement attributes."""
    compute_time: bool
    mac_operations: bool
    communication: bool
    forgetting: bool
    forward_transfer: bool
    backward_transfer: bool
    openness: bool
    parameters: bool
    memory: bool
    stored_data: bool
    generated_data: bool
    optimization_steps: bool
    per_task_metric: bool
    task_order: bool
    data_per_task: bool

    def __iter__(self):
        """
        Defines the iteration order. This needs to be the same order as defined in the
        cleva_template.tex file.
        """
        for item in [self.parameters, self.compute_time, self.mac_operations, self.communication, self.forgetting, self.forward_transfer, self.backward_transfer, self.openness, self.data_per_task, self.task_order, self.per_task_metric, self.optimization_steps, self.generated_data, self.stored_data, self.memory]:
            yield item

@dataclass
class CompassEntry:
    """Compass entry containing color, label, and attributes."""
    color: str
    label: str
    inner_level: InnerLevel
    outer_level: OuterLevel

# Create sample entries
inner_level_1 = InnerLevel(
    multiple_models=1,
    federated=1,
    online=0,
    open_world=1,
    multiple_modalities=0,
    active_data_query=1,
    task_order_discovery=0,
    task_agnostic=1,
    episodic_memory=1,
    generative=0,
    uncertainty=1
)

outer_level_1 = OuterLevel(
    compute_time=True,
    mac_operations=True,
    communication=False,
    forgetting=True,
    forward_transfer=False,
    backward_transfer=True,
    openness=False,
    parameters=True,
    memory=True,
    stored_data=False,
    generated_data=True,
    optimization_steps=False,
    per_task_metric=True,
    task_order=False,
    data_per_task=True
)

entry_1 = CompassEntry(
    color='blue',
    label='Method A',
    inner_level=inner_level_1,
    outer_level=outer_level_1
)

inner_level_2 = InnerLevel(
    multiple_models=0,
    federated=0,
    online=1,
    open_world=0,
    multiple_modalities=1,
    active_data_query=0,
    task_order_discovery=1,
    task_agnostic=0,
    episodic_memory=0,
    generative=1,
    uncertainty=0
)

outer_level_2 = OuterLevel(
    compute_time=False,
    mac_operations=True,
    communication=True,
    forgetting=False,
    forward_transfer=True,
    backward_transfer=False,
    openness=True,
    parameters=False,
    memory=False,
    stored_data=True,
    generated_data=False,
    optimization_steps=True,
    per_task_metric=False,
    task_order=True,
    data_per_task=False
)

entry_2 = CompassEntry(
    color='green',
    label='Method B',
    inner_level=inner_level_2,
    outer_level=outer_level_2
)

entries = [entry_1, entry_2]

# Create a minimal template string
template_content = """
%-$NUMBER-OF-METHODS$
%-$INNER-CIRCLE$
%-$OUTER-CIRCLE$
%-$LEGEND$
"""

# Create a temporary template file
import tempfile
import os

temp_dir = tempfile.gettempdir()
template_path = os.path.join(temp_dir, 'test_template.tex')
with open(template_path, 'w') as f:
    f.write(template_content)

ARGS = {
    'template_path': template_path,
    'entries': entries
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
insert_legend Ran
ran; our contract draft was wrong, not the code
ml-research/CLEVA-Compass/create_compass.py
code served (permissive licence) · get_code("3ff69caeb20a8e83")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import math

class LegendEntry:
    def __init__(self, color, label):
        self.color = color
        self.label = label

# Create test entries with different colors and labels
entries = [
    LegendEntry('magenta', 'Entry 1'),
    LegendEntry('green', 'Entry 2'),
    LegendEntry('blue', 'Entry 3'),
    LegendEntry('orange', 'Entry 4'),
]

template = "Some template text\n%-$LEGEND$\nMore template text"

ARGS = {
    'template': template,
    'entries': entries,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
insert_number_of_methods Ran
ran; our contract draft was wrong, not the code
ml-research/CLEVA-Compass/create_compass.py
code served (permissive licence) · get_code("977ff51a91af0715")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np
from dataclasses import dataclass
from typing import List

@dataclass
class InnerLevel:
    """Inner CLEVA-Compass level with method attributes."""
    multiple_models: int
    federated: int
    online: int
    open_world: int
    multiple_modalities: int
    active_data_query: int
    task_order_discovery: int
    task_agnostic: int
    episodic_memory: int
    generative: int
    uncertainty: int

    def __iter__(self):
        """
        Defines the iteration order. This needs to be the same order as defined in the
        cleva_template.tex file.
        """
        for item in [self.multiple_models, self.federated, self.online, self.open_world, self.multiple_modalities, self.active_data_query, self.task_order_discovery, self.task_agnostic, self.episodic_memory, self.generative, self.uncertainty]:
            yield item

@dataclass
class OuterLevel:
    """Outer CLEVA-Compass level with measurement attributes."""
    compute_time: bool
    mac_operations: bool
    communication: bool
    forgetting: bool
    forward_transfer: bool
    backward_transfer: bool
    openness: bool
    parameters: bool
    memory: bool
    stored_data: bool
    generated_data: bool
    optimization_steps: bool
    per_task_metric: bool
    task_order: bool
    data_per_task: bool

    def __iter__(self):
        """
        Defines the iteration order. This needs to be the same order as defined in the
        cleva_template.tex file.
        """
        for item in [self.parameters, self.compute_time, self.mac_operations, self.communication, self.forgetting, self.forward_transfer, self.backward_transfer, self.openness, self.data_per_task, self.task_order, self.per_task_metric, self.optimization_steps, self.generated_data, self.stored_data, self.memory]:
            yield item

@dataclass
class CompassEntry:
    """Compass entry containing color, label, and attributes."""
    color: str
    label: str
    inner_level: InnerLevel
    outer_level: OuterLevel

def insert_number_of_methods(template, entries: List[CompassEntry]):
    """Insert number of methods as newcommand \\M."""
    n_methods_str = '\\newcommand{\\M}{' + str(len(entries)) + '}'
    template = template.replace('%-$NUMBER-OF-METHODS$', n_methods_str)
    return template


# Create sample InnerLevel instance
inner_level_1 = InnerLevel(
    multiple_models=1,
    federated=0,
    online=1,
    open_world=0,
    multiple_modalities=1,
    active_data_query=0,
    task_order_discovery=1,
    task_agnostic=0,
    episodic_memory=1,
    generative=0,
    uncertainty=1
)

# Create sample OuterLevel instance
outer_level_1 = OuterLevel(
    compute_time=True,
    mac_operations=False,
    communication=True,
    forgetting=False,
    forward_transfer=True,
    backward_transfer=False,
    openness=True,
    parameters=False,
    memory=True,
    stored_data=False,
    generated_data=True,
    optimization_steps=False,
    per_task_metric=True,
    task_order=False,
    data_per_task=True
)

# Create sample CompassEntry instances
entry_1 = CompassEntry(
    color="red",
    label="Method1",
    inner_level=inner_level_1,
    outer_level=outer_level_1
)

entry_2 = CompassEntry(
    color="blue",
    label="Method2",
    inner_level=inner_level_1,
    outer_level=outer_level_1
)

# Create list of entries
entries = [entry_1, entry_2]

# Create template string with placeholder
template = "%-$NUMBER-OF-METHODS$ Some LaTeX content here"

ARGS = {
    "template": template,
    "entries": entries
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
mapcolor Ran
ran; our contract draft was wrong, not the code
ml-research/CLEVA-Compass/create_compass.py
code served (permissive licence) · get_code("302927553501b70e")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

# Setup code for mapcolor function

ARGS = {
    'color': 'magenta'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
update_background Ran
ran; our contract draft was wrong, not the code
ml-research/CLEVA-Compass/cleva_gui.py
code served (permissive licence) · get_code("982ca0b928b57cd9")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned Image. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

from PIL import Image

# Create a small test image with some white pixels and some non-white pixels
img = Image.new('RGB', (4, 3))
pixdata = img.load()

# Set some pixels to white (255, 255, 255) and others to different colors
pixdata[0, 0] = (255, 255, 255)  # white - should be changed
pixdata[1, 0] = (100, 100, 100)  # gray - should stay the same
pixdata[2, 0] = (255, 255, 255)  # white - should be changed
pixdata[3, 0] = (200, 50, 75)    # colored - should stay the same
pixdata[0, 1] = (255, 255, 255)  # white - should be changed
pixdata[1, 1] = (0, 0, 0)        # black - should stay the same
pixdata[2, 1] = (255, 255, 255)  # white - should be changed
pixdata[3, 1] = (150, 150, 150)  # gray - should stay the same
pixdata[0, 2] = (255, 255, 255)  # white - should be changed
pixdata[1, 2] = (50, 100, 200)   # blue - should stay the same
pixdata[2, 2] = (255, 255, 255)  # white - should be changed
pixdata[3, 2] = (75, 75, 75)     # gray - should stay the same

ARGS = {
    'img': img
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
PIL — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
insert_inner_level Not yet run
unverified (harvested, no recorded run)
ml-research/CLEVA-Compass/create_compass.py
code served (permissive licence) · get_code("e93ca7234a4de368")
How it ran

no invocation recorded.

insert_outer_level Not yet run
unverified (harvested, no recorded run)
ml-research/CLEVA-Compass/create_compass.py
code served (permissive licence) · get_code("a756bab4d915ae92")
How it ran

no invocation recorded.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 10 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

What is the state of the art in continual machine learning? Although a natural question for predominant static benchmarks, the notion to train systems in a lifelong manner entails a plethora of additional challenges with respect to set-up and evaluation. The latter have recently sparked a growing amount of critiques on prominent algorithm-centric perspectives and evaluation protocols being too narrow, resulting in several attempts at constructing guidelines in favor of specific desiderata or arguing against the validity of prevalent assumptions. In this work, we depart from this mindset and argue that the goal of a precise formulation of desiderata is an ill-posed one, as diverse applications may always warrant distinct scenarios. Instead, we introduce the Continual Learning EValuation Assessment Compass: the CLEVA-Compass. The compass provides the visual means to both identify how approaches are practically reported and how works can simultaneously be contextualized in the broader literature landscape. In addition to promoting compact specification in the spirit of recent replication trends, it thus provides an intuitive chart to understand the priorities of individual systems, where they resemble each other, and what elements are missing towards a fair comparison.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2110.03331")
get_code_for_paper("2110.03331")
have("2110.03331")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2110.03331.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2110.03331.svg)](https://syntology.ai/paper/2110.03331)

Connect an agent — have() is free.