SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2003.04300 · 2020

Learning Discrete State Abstractions With Deep Variational Inference

arXiv · PDF · Open in the Atlas

Code that ran

5 of 7 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 5 have no contract verdict; 0 failures of our instrument; 2 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/2003.04300.json · badge: /api/ran/2003.04300.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
ondrejba/discrete_abstractions canonical (classifier label)
classifier basis: census
5 ran · 2 no recorded run
FunctionStatusWhere it lives
get_perplexities Ran
ran on a synthesized input
ondrejba/discrete_abstractions/evaluate.py
code served (permissive licence) · get_code("bb49644247032715")
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; no output shape is stated. It ran in syntology-harvest-sandbox:v5-deps (sha256:6a292aa1aed1) at 2026-09-27T00:55:05Z. 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

# Analyze the function:
# - get_perplexities takes log_distribution as input
# - It performs numpy operations: division, exponentiation, sum along axis=1
# - log_distribution should be a 2D array (batch_size, vocab_size)
# - axis=1 suggests rows are samples, columns are vocabulary/distribution values

# Create a small log-probability distribution
# Shape: (batch_size=2, vocab_size=4)
# Values should be log probabilities (negative numbers, sum to log(1) = 0 when exponentiated)
log_distribution = np.array([
    [-1.386, -2.303, -2.996, -3.507],  # batch 0: log probs over 4 items
    [-0.693, -1.609, -2.303, -3.219],  # batch 1: log probs over 4 items
], dtype=np.float32)

ARGS = {
    "log_distribution": log_distribution
}
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 — no output shape is stated: no later run recorded an output shape for this sample (observed_output and metamorphic_properties carry none; shape_contract and output_contract record ndim only)
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 syntology-harvest-sandbox:v5-deps (sha256:6a292aa1aed1); 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
multiplex_hand_state Ran
ran on a synthesized input
ondrejba/discrete_abstractions/vis_utils.py
code served (permissive licence) · get_code("7fa38275d1aa2fe1")
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 (6, 8) 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)

# Construct a state tuple containing depth image and hand state
# The function expects:
# - state[0]: a depth image (tensor that supports indexing and assignment)
# - state[1]: hand_state (a numeric value to check if > 0)

depth_image = torch.randn(6, 8)  # Small depth image with distinct dimensions
hand_state_value = 1  # Positive value to trigger the if condition

state = (depth_image, hand_state_value)

# paint_value is a scalar depth value to paint into the depth image
paint_value = 0.5

ARGS = {
    "state": state,
    "paint_value": paint_value
}
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 (6, 8) — shape (6, 8) 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
states_cutoff Ran
ran on a synthesized input
ondrejba/discrete_abstractions/model/utils.py
code served (permissive licence) · get_code("f1a53d06a5849890")
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 numpy as np

# Construct valid arguments for states_cutoff
num_blocks = 5
batch_size = 20

# assignments: array of block indices for each sample
# Using values in range [0, num_blocks) to match blocks that will be counted
assignments = np.array([0, 0, 1, 1, 2, 2, 2, 3, 4, 4, 
                        0, 1, 2, 3, 4, 0, 1, 2, 3, 4], dtype=np.int32)

# r_hat: shape (batch_size, num_blocks) - response matrix
r_hat = np.random.randn(batch_size, num_blocks).astype(np.float32)

# t_hat: shape (batch_size, num_blocks, num_blocks) - transition matrix
t_hat = np.random.randn(batch_size, num_blocks, num_blocks).astype(np.float32)

# cutoff_fraction: fraction threshold for filtering
cutoff_fraction = 0.2

ARGS = {
    "assignments": assignments,
    "r_hat": r_hat,
    "t_hat": t_hat,
    "num_blocks": num_blocks,
    "cutoff_fraction": cutoff_fraction
}
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 — 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
transform_embeddings Ran
ran on a synthesized input
ondrejba/discrete_abstractions/model/utils.py
code served (permissive licence) · get_code("ac62be94abe88744")
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 2, dtype family float; no output shape is stated. It ran in syntology-harvest-sandbox:v5-deps (sha256:6a292aa1aed1) at 2026-09-27T00:55:05Z. 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

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

# Create small test data
batch_size = 2
embedding_dim = 4
num_actions = 3
transform_dim = 5

# embeddings: shape (batch_size, embedding_dim)
embeddings = np.random.randn(batch_size, embedding_dim).astype(np.float32)

# t_hat: shape (num_actions, embedding_dim, transform_dim)
# This is indexed by actions, so needs num_actions as first dimension
t_hat = np.random.randn(num_actions, embedding_dim, transform_dim).astype(np.float32)

# actions: shape (batch_size,) with indices into the first dimension of t_hat
actions = np.array([0, 1], dtype=np.int64)

ARGS = {
    "embeddings": embeddings,
    "t_hat": t_hat,
    "actions": actions,
}
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 2, dtype family float — no output shape is stated: no later run recorded an output shape for this sample (observed_output and metamorphic_properties carry none; shape_contract and output_contract record ndim only)
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 syntology-harvest-sandbox:v5-deps (sha256:6a292aa1aed1); 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
unpack_states Ran
ran on a synthesized input
ondrejba/discrete_abstractions/vis_utils.py
code served (permissive licence) · get_code("154b245eb25ffb95")
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 numpy as np

# Create sample states
# Each state is a tuple of (depth_image, hand_state)
# depth_image: 2D array of depth values (batch of 2, spatial dims 4x3)
# hand_state: 1D array of hand joint positions (batch of 2, 5 joints)

states = [
    (np.random.rand(4, 3).astype(np.float32), np.random.randint(0, 10, size=5).astype(np.int32)),
    (np.random.rand(4, 3).astype(np.float32), np.random.randint(0, 10, size=5).astype(np.int32)),
]

ARGS = {
    "states": states,
}
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 — 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
multiplex_action Not yet run
unverified (harvested, no recorded run)
ondrejba/discrete_abstractions/vis_utils.py
code served (permissive licence) · get_code("cbab59bd4ee79f50")
How it ran

no invocation recorded.

transform_and_plot_embeddings Not yet run
unverified (harvested, no recorded run)
ondrejba/discrete_abstractions/model/utils.py
code served (permissive licence) · get_code("17df876bed242d2e")
How it ran

no invocation recorded.

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 carry a function identity and 0 do not. Of the 7 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 7 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

Abstraction is crucial for effective sequential decision making in domains with large state spaces. In this work, we propose an information bottleneck method for learning approximate bisimulations, a type of state abstraction. We use a deep neural encoder to map states onto continuous embeddings. We map these embeddings onto a discrete representation using an action-conditioned hidden Markov model, which is trained end-to-end with the neural network. Our method is suited for environments with high-dimensional states and learns from a stream of experience collected by an agent acting in a Markov decision process. Through this learned discrete abstract model, we can efficiently plan for unseen goals in a multi-goal Reinforcement Learning setting. We test our method in simplified robotic manipulation domains with image states. We also compare it against previous model-based approaches to finding bisimulations in discrete grid-world-like environments. Source code is available at https://github.com/ondrejba/discrete_abstractions.

For agents

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

get_harvested_code_for_paper("2003.04300")
get_code_for_paper("2003.04300")
have("2003.04300")

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

curl https://syntology.ai/api/ran/2003.04300.json

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

[![Syntology run record](https://syntology.ai/api/ran/2003.04300.svg)](https://syntology.ai/paper/2003.04300)

Connect an agent — have() is free.