SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2107.08295 · ICML · 2022

Communicating via Markov Decision Processes

J Zico Kolter, Shimon Whiteson, Jakob Foerster, Philip Torr, Samuel Sokota, Luisa Zintgraf, Maximilian Igl, Christian De Witt, Martin Strohmeier

arXiv · PDF · Open in the Atlas

Code that ran

4 of 5 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 0 have no contract verdict; 4 failures of our instrument; 1 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/2107.08295.json · badge: /api/ran/2107.08295.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
schroederdewitt/meme canonical (classifier label)
classifier basis: classifier_2025
4 ran · 1 no recorded run
FunctionStatusWhere it lives
Lemma3 Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
schroederdewitt/meme/src/minimum_entropy_utils.py
code served (permissive licence) · get_code("1d1b5ec53efcb7b7")
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 1, 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 torch as th

th.manual_seed(0)

# Analyze function usage:
# - z: scalar, used in comparison and arithmetic (z - z_r)
# - x: scalar, used in comparison (x >= 0) and arithmetic (A.sum() + x, A[:k].sum() - x)
# - A: tensor, used with indexing A[:k], A[A < 0.0], .sum() operations
# - i: integer, used as starting index k = i
# - j: integer, used in comparison k > j

# Create valid arguments that satisfy all assertions:
# 1. x >= 0
# 2. A must be non-negative (A[A < 0.0].sum() == 0)
# 3. A.sum() + x >= z

z = th.tensor(15.0, dtype=th.float64)
x = 5.0
A = th.tensor([2.0, 3.0, 4.0, 6.0], dtype=th.float64)  # all non-negative
i = 0
j = 3

# Verify constraints:
# x >= 0: 5.0 >= 0 ✓
# A non-negative: all values are positive ✓
# A.sum() + x >= z: (2+3+4+6) + 5 = 20 >= 15 ✓

ARGS = {
    "z": z,
    "x": x,
    "A": A,
    "i": i,
    "j": j,
}
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 tuple, ndim 1, 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
greatest_lower_bound Ran
ran; our contract draft was wrong, not the code
schroederdewitt/meme/src/minimum_entropy_utils.py
code served (permissive licence) · get_code("1171ee49436edbbe")
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 as th

th.manual_seed(0)

# Create two tensors with matching shapes
# batch_size=2, dim=4
p = th.randn(2, 4)
q = th.randn(2, 4)

# Ensure both tensors have non-negative values since the function
# appears to work with cumulative sums and minimums
p = th.abs(p)
q = th.abs(q)

ARGS = {
    'p': p,
    'q': q,
}
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 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
load_image_to_token Ran
ran; our contract draft was wrong, not the code
schroederdewitt/meme/src/minimum_entropy.py
code served (permissive licence) · get_code("5386a45e20e1af97")
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 list, ndim 1, dtype family int; 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
from matplotlib import pyplot as plt
from PIL import Image
import io

# Create a small test image in memory
# The function expects an image where it reads the first channel
# and flattens it, then processes it based on chunks
# We need to ensure the total number of pixels matches the sum of bitlengths

# For chunks = [4, 8, 16], we need:
# bitlen for 4 = log2(4) = 2
# bitlen for 8 = log2(8) = 3
# bitlen for 16 = log2(16) = 4
# Total = 2 + 3 + 4 = 9 pixels needed

# Create a 3x3 image (9 pixels) with RGB channels
img_array = np.random.randint(0, 256, (3, 3, 3), dtype=np.uint8)

# Save to bytes buffer
buf = io.BytesIO()
img = Image.fromarray(img_array)
img.save(buf, format='PNG')
buf.seek(0)

# Read it back to get a proper PNG file path simulation
# For this setup, we'll create a temporary file-like approach
# Actually, let's just create the image array directly and mock the path

# Create a simple RGB image as a numpy array and save it temporarily
import tempfile
import os

temp_dir = tempfile.gettempdir()
test_image_path = os.path.join(temp_dir, "test_image_for_token.png")

# Create and save a test image
test_img = np.random.randint(0, 256, (3, 3, 3), dtype=np.uint8)
Image.fromarray(test_img).save(test_image_path)

# Define chunks such that sum of log2(c) equals total pixels
# For a 3x3 image = 9 pixels
# We can use chunks = [4, 8, 16] which gives log2 values of 2, 3, 4 = 9 total
chunks = [4, 8, 16]

ARGS = {
    "path": test_image_path,
    "chunks": chunks
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
PIL, matplotlib, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family int — a list 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
minimum_entropy_coupling Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
schroederdewitt/meme/src/minimum_entropy_utils.py
code served (permissive licence) · get_code("88efedce36b9b019")
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 dict. 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 as th

th.manual_seed(0)

# Create two valid probability distributions
# p and q must be 1D tensors that sum to 1 and are non-negative
p = th.tensor([0.5, 0.3, 0.15, 0.05], dtype=th.float64)
q = th.tensor([0.25, 0.25, 0.25, 0.25], dtype=th.float64)

ARGS = {
    'p': p,
    'q': q,
    'atol': 1e-06,
    'rtol': None,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
get_free_device_memory Not yet run
unverified (harvested, no recorded run)
schroederdewitt/meme/src/train_ff.py
code served (permissive licence) · get_code("e50631fd4dc5963e")
How it ran

no invocation recorded. this sample carries JSON test cases instead of a setup block; get_reference_implementation serves them with the level they back.

Also present in other papers' code

Of 5 samples lifted from this paper's linked repositories, 5 carry a function identity and 0 do not. Of the 5 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 5 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

We consider the problem of communicating exogenous information by means of Markov decision process trajectories. This setting, which we call a Markov coding game (MCG), generalizes both source coding and a large class of referential games. MCGs also isolate a problem that is important in decentralized control settings in which cheap-talk is not available-namely, they require balancing communication with the associated cost of communicating. We contribute a theoretically grounded approach to MCGs based on maximum entropy reinforcement learning and minimum entropy coupling that we call MEME. Due to recent breakthroughs in approximation algorithms for minimum entropy coupling, MEME is not merely a theoretical algorithm, but can be applied to practical settings. Empirically, we show both that MEME is able to outperform a strong baseline on small MCGs and that MEME is able to achieve strong performance on extremely large MCGs. To the latter point, we demonstrate that MEME is able to losslessly communicate binary images via trajectories of Cartpole and Pong, while simultaneously achieving the maximal or near maximal expected returns, and that it is even capable of performing well in the presence of actuator noise.

For agents

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

get_harvested_code_for_paper("2107.08295")
get_code_for_paper("2107.08295")
have("2107.08295")

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

curl https://syntology.ai/api/ran/2107.08295.json

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

[![Syntology run record](https://syntology.ai/api/ran/2107.08295.svg)](https://syntology.ai/paper/2107.08295)

Connect an agent — have() is free.