SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2304.13892 · ICML · 2023

Discovering Object-Centric Generalized Value Functions From Pixels

Khimya Khetarpal, Gopeshh Subbaraj, Somjit Nath, Samira Ebrahimi Kahou

arXiv · PDF · Open in the Atlas

Code that ran

14 of 18 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 10 have no contract verdict; 4 failures of our instrument; 4 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; 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/2304.13892.json · badge: /api/ran/2304.13892.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
Somjit77/oc_gvfs canonical (classifier label)
classifier basis: census
10 ran · 1 no recorded run
somjit77/oc_gvfs canonical (classifier label)
classifier basis: classifier_2025
4 ran · 3 no recorded run
FunctionStatusWhere it lives
build_grid Ran
ran; our contract draft was wrong, not the code
somjit77/oc_gvfs/slot_attention/model_jax.py
code served (permissive licence) · get_code("d42ce6eadbd4a112")
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 ArrayImpl, ndim 4, dtype family float, shape (1, 4, 3, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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 jax.numpy as jnp

# The function build_grid takes a single parameter: resolution
# resolution should be a sequence of integers specifying grid dimensions
# Looking at the function body:
# - resolution is indexed with [0] and [1], so it needs at least 2 elements
# - resolution values are passed to jnp.linspace as num parameter
# - The result is reshaped using resolution[0] and resolution[1]

# Create a small resolution tuple with distinct dimensions for probing
resolution = (4, 3)

ARGS = {
    "resolution": resolution
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 4, dtype family float, shape (1, 4, 3, 4) — shape (1, 4, 3, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
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
build_network Ran
ran on a synthesized input
Somjit77/oc_gvfs/agent.py
code served (permissive licence) · get_code("4f1d87d8590a8749")
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 Transformed. 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 haiku as hk
import jax
import jax.numpy as jnp

ARGS = {
    'num_actions': 4,
    'num_gvfs': 2,
    'hidden_units': [32, 16],
    'algorithm': 'esp',
    'use_action_values': True,
    'use_concatanation': True,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
haiku, jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Transformed — a Transformed output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Transformed output the recorded ndim 1 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
build_question_network Ran
ran on a synthesized input
Somjit77/oc_gvfs/agent.py
code served (permissive licence) · get_code("4eca72bd8d6a28b0")
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 Transformed. 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 jax
import jax.numpy as jnp

ARGS = {
    'num_gvfs': 3,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
haiku, jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Transformed — a Transformed output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Transformed output the recorded ndim 1 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
newest Ran
ran on a synthesized input
Somjit77/oc_gvfs/plot_curves.py
code served (permissive licence) · get_code("c2586f375dd0aef5")
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 os
import tempfile
import shutil

# Create a temporary directory with some test files
temp_dir = tempfile.mkdtemp()

# Create a few test files with different modification times
file1 = os.path.join(temp_dir, "file1.txt")
file2 = os.path.join(temp_dir, "file2.txt")
file3 = os.path.join(temp_dir, "file3.txt")

# Write files with slight delays to ensure different ctimes
with open(file1, 'w') as f:
    f.write("test1")
    
with open(file2, 'w') as f:
    f.write("test2")
    
with open(file3, 'w') as f:
    f.write("test3")

ARGS = {
    'path': temp_dir
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
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
newest Ran
ran on a synthesized input
Somjit77/oc_gvfs/slot_attention/sa_utils.py
code served (permissive licence) · get_code("bd938d576acc121a")
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 os
import tempfile
import shutil

# Create a temporary directory with some test files
temp_dir = tempfile.mkdtemp()

# Create a few test files with different modification times
test_file_1 = os.path.join(temp_dir, "file1.txt")
test_file_2 = os.path.join(temp_dir, "file2.txt")
test_file_3 = os.path.join(temp_dir, "file3.txt")

# Write files in sequence to ensure different ctimes
with open(test_file_1, 'w') as f:
    f.write("content1")

with open(test_file_2, 'w') as f:
    f.write("content2")

with open(test_file_3, 'w') as f:
    f.write("content3")

# Argument dictionary for the newest function
ARGS = {
    "path": temp_dir
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
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
pre_process_batch Ran
ran on a synthesized input
Somjit77/oc_gvfs/slot_attention/sa_utils.py
code served (permissive licence) · get_code("c73955955ab59b07")
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 ArrayImpl, ndim 4, dtype family float, shape (2, 48, 48, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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 jax
import jax.numpy as jnp

# Create a small batch of images with shape (batch, height, width, channels)
# batch=2, height=64, width=64, channels=3
batch = jnp.ones((2, 64, 64, 3), dtype=jnp.float32)

# resolution parameter - tuple of (height, width) for resizing
resolution = (48, 48)

ARGS = {
    "batch": batch,
    "resolution": resolution,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 4, dtype family float, shape (2, 48, 48, 3) — shape (2, 48, 48, 3) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
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
renormalize Ran
ran on a synthesized input
Somjit77/oc_gvfs/slot_attention/sa_utils.py
code served (permissive licence) · get_code("58e7e6adb605babe")
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 Tensor, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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

torch.manual_seed(0)

# The function renormalize takes a single parameter x
# and performs arithmetic operations on it (division and addition)
# It expects x to be a numeric tensor or array

# Create a small tensor in the range [-1, 1] to test the renormalization
x = torch.linspace(-1, 1, 8).reshape(2, 4)

ARGS = {
    "x": x
}
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 Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
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
running_mean Ran
ran on a synthesized input
Somjit77/oc_gvfs/plot_curves.py
code served (permissive licence) · get_code("85a38e4dd9181aaf")
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 ndarray, ndim 1, dtype family float, shape (6,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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 numpy as np

# The function takes X (array-like) and N (integer)
# X should be a 1D array of numbers
# N should be an integer representing the window size

X = np.array([1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0])
N = 3

ARGS = {
    "X": X,
    "N": N
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family float, shape (6,) — shape (6,) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
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
spatial_broadcast Ran
ran; our contract draft was wrong, not the code
somjit77/oc_gvfs/slot_attention/model_jax.py
code served (permissive licence) · get_code("85d66ddd34119a7c")
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 ArrayImpl, ndim 4, dtype family float, shape (6, 5, 6, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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 jax.numpy as jnp

# Create test inputs for spatial_broadcast
# The function expects:
# - slots: a JAX array with shape [..., num_slots, feature_dim]
# - resolution: a tuple/list [height, width]

# Create a small slots array with shape (batch=2, num_slots=3, feature_dim=4)
slots = jnp.ones((2, 3, 4))

# Create a resolution tuple
resolution = (5, 6)

ARGS = {
    "slots": slots,
    "resolution": resolution,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 4, dtype family float, shape (6, 5, 6, 4) — shape (6, 5, 6, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
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
spatial_flatten Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
somjit77/oc_gvfs/slot_attention/model_jax.py
code served (permissive licence) · get_code("ca839bba3cb3de30")
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 ArrayImpl, ndim 3, dtype family float, shape (2, 12, 5) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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 jax.numpy as jnp

# The function spatial_flatten takes a single argument x
# It expects x to be an array-like object with shape attributes
# and supports jnp.reshape and indexing operations

# Create a small test array with shape (batch, height, width, channels)
# Using distinct sizes: batch=2, height=3, width=4, channels=5
x = jnp.ones((2, 3, 4, 5))

ARGS = {
    "x": x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 3, dtype family float, shape (2, 12, 5) — shape (2, 12, 5) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
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
td_error Ran
ran on a synthesized input
Somjit77/oc_gvfs/utils.py
code served (permissive licence) · get_code("5daf1e1813a04ece")
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 ArrayImpl, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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 jax
import jax.numpy as jnp
import numpy as np

# Set seed for reproducibility
np.random.seed(0)

# Create small JAX arrays for the function parameters
# q_tm1: Q-values at time t-1, shape (num_actions,)
q_tm1 = jnp.array([0.1, 0.2, 0.3, 0.4], dtype=jnp.float32)

# a_tm1: Action index at time t-1, scalar int
a_tm1 = jnp.array(1, dtype=jnp.int32)

# cumulant: Reward or return signal, scalar
cumulant = jnp.array(1.5, dtype=jnp.float32)

# discount_t: Discount factor, scalar
discount_t = jnp.array(0.99, dtype=jnp.float32)

# q_t_val: Q-values at time t, shape (num_actions,)
q_t_val = jnp.array([0.2, 0.3, 0.25, 0.35], dtype=jnp.float32)

# q_t_select: Q-values used for action selection at time t, shape (num_actions,)
q_t_select = jnp.array([0.15, 0.25, 0.2, 0.3], dtype=jnp.float32)

ARGS = {
    "q_tm1": q_tm1,
    "a_tm1": a_tm1,
    "cumulant": cumulant,
    "discount_t": discount_t,
    "q_t_val": q_t_val,
    "q_t_select": q_t_select,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
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
td_error_state Ran
ran on a synthesized input
Somjit77/oc_gvfs/utils.py
code served (permissive licence) · get_code("5dcbe772a2a9f6b7")
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 ArrayImpl, ndim 1, dtype family float, shape (4,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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 jax
import jax.numpy as jnp

# The function expects JAX arrays (or array-like objects)
# v_tm1: value at time t-1
# cumulant: immediate reward/cumulant
# discount_t: discount factor at time t
# v_t_val: value at time t

ARGS = {
    "v_tm1": jnp.array([0.5, -0.3, 0.2, 0.1]),  # Value estimates at t-1
    "cumulant": jnp.array([1.0, 0.5, -0.2, 0.8]),  # Immediate cumulants
    "discount_t": jnp.array([0.99, 0.99, 0.99, 0.99]),  # Discount factors
    "v_t_val": jnp.array([0.3, 0.4, 0.1, 0.2])  # Value estimates at t
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ArrayImpl, ndim 1, dtype family float, shape (4,) — shape (4,) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
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
tolerant_mean Ran
ran on a synthesized input
Somjit77/oc_gvfs/plot_curves.py
code served (permissive licence) · get_code("a52c31e8052f488b")
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 tuple, ndim 2, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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 numpy as np

# Construct arguments for tolerant_mean
# The function expects arrs: a list of arrays of potentially different lengths

# Create arrays of different lengths to test the tolerant_mean function
arr1 = np.array([1.0, 2.0, 3.0, 4.0])
arr2 = np.array([5.0, 6.0])
arr3 = np.array([7.0, 8.0, 9.0])

arrs = [arr1, arr2, arr3]

ARGS = {
    'arrs': arrs
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 2, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
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
unstack_and_split Ran
ran; our contract draft was wrong, not the code
somjit77/oc_gvfs/slot_attention/model_jax.py
code served (permissive licence) · get_code("e48ab0090f3934eb")
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 tuple. 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 jax.numpy as jnp
import numpy as np

# Create input tensor x
# The function expects x to be reshaped to [batch_size, -1, ...] and then split on last axis
# It splits into 4 parts (c1, c2, c3, masks), so last dimension should be divisible by 4
# Let's create shape: (batch_size=2, height=4, width=3, channels=4)
# This will reshape to [2, -1, 4] where -1 = 4*3 = 12
# Then split on last axis gives 4 tensors of shape [2, 12, 1]

batch_size = 2
height = 4
width = 3
total_channels = 4  # Will be split into c1, c2, c3, masks

x = jnp.ones((batch_size, height, width, total_channels), dtype=jnp.float32)

ARGS = {
    "x": x,
    "batch_size": batch_size,
    "num_channels": 3
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is 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
SlotAttention Not yet run
unverified (harvested, no recorded run)
somjit77/oc_gvfs/slot_attention/model_jax.py
code served (permissive licence) · get_code("2f149edc4ef4e955")
How it ran

no invocation recorded.

SlotAttentionAutoEncoder Not yet run
unverified (harvested, no recorded run)
somjit77/oc_gvfs/slot_attention/model_jax.py
code served (permissive licence) · get_code("af260099be8d330b")
How it ran

no invocation recorded.

SoftPositionEmbed Not yet run
unverified (harvested, no recorded run)
somjit77/oc_gvfs/slot_attention/model_jax.py
code served (permissive licence) · get_code("9d993e405d2762ce")
How it ran

no invocation recorded.

get_cumulants Not yet run
unverified (harvested, no recorded run)
Somjit77/oc_gvfs/utils.py
code served (permissive licence) · get_code("870f819ffac4dd2c")
How it ran

no invocation recorded.

Also present in other papers' code

Of 18 samples lifted from this paper's linked repositories, 18 carry a function identity and 0 do not. Of the 17 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 17 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

Deep Reinforcement Learning has shown significant progress in extracting useful representations from high-dimensional inputs albeit using handcrafted auxiliary tasks and pseudo rewards. Automatically learning such representations in an object-centric manner geared towards control and fast adaptation remains an open research problem. In this paper, we introduce a method that tries to discover meaningful features from objects, translating them to temporally coherent 'question' functions and leveraging the subsequent learned general value functions for control. We compare our approach with state-of-the-art techniques alongside other ablations and show competitive performance in both stationary and non-stationary settings. Finally, we also investigate the discovered general value functions and through qualitative analysis show that the learned representations are not only interpretable but also, centered around objects that are invariant to changes across tasks facilitating fast adaptation.

For agents

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

get_harvested_code_for_paper("2304.13892")
get_code_for_paper("2304.13892")
have("2304.13892")

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

curl https://syntology.ai/api/ran/2304.13892.json

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

[![Syntology run record](https://syntology.ai/api/ran/2304.13892.svg)](https://syntology.ai/paper/2304.13892)

Connect an agent — have() is free.