SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2406.04327 · ACL · 2024

Causal Estimation of Memorisation Profiles

Thomas Hofmann, Andreas Vlachos, Tiago Pimentel, Clara Meister, Pietro Lesci

arXiv · PDF · Open in the Atlas

Code that ran

3 of 3 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; 1 honoured the contract we drafted, 0 violated it and 1 has no contract verdict; 1 failure of our instrument; 0 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/2406.04327.json · badge: /api/ran/2406.04327.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
pietrolesci/memorisation-profiles canonical (classifier label)
classifier basis: census
3 ran · 0 no recorded run
FunctionStatusWhere it lives
convert_json_to_pandas Ran
ran on a synthesized input
pietrolesci/memorisation-profiles/src/utilities.py
code served (permissive licence) · get_code("6d15dd7c41fd0a5e")
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 DataFrame. 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 pandas as pd
import json as json_module
import tempfile
import os

# Create a temporary JSON file with valid data
temp_dir = tempfile.mkdtemp()
json_file_path = os.path.join(temp_dir, "test_data.json")

# Write valid JSON data to the file
test_data = [
    {"id": 1, "name": "Alice", "value": 10.5},
    {"id": 2, "name": "Bob", "value": 20.3},
    {"id": 3, "name": "Charlie", "value": 15.7}
]

with open(json_file_path, 'w') as f:
    for item in test_data:
        f.write(json_module.dumps(item) + '\n')

ARGS = {
    'path': json_file_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
datasets, pandas, pyarrow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 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
flatten Ran
ran, honoured the contract we drafted
pietrolesci/memorisation-profiles/src/utilities.py
code served (permissive licence) · get_code("d2d591b2619db1bd")
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 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 torch
import numpy as np

torch.manual_seed(0)

# The function flatten takes a nested iterable x and flattens it by one level
# It iterates over x (outer loop: for j in x) and then iterates over each element j (inner loop: for i in j)
# So x should be a list of iterables

# Create a simple nested structure - list of lists
nested_list = [
    [1, 2, 3],
    [4, 5],
    [6, 7, 8, 9]
]

ARGS = {
    "x": nested_list
}
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 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
ld_to_dl Ran
ran; our contract draft was wrong, not the code
pietrolesci/memorisation-profiles/src/utilities.py
code served (permissive licence) · get_code("a0a0c7a71d2fa750")
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 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
import numpy as np

torch.manual_seed(0)

# Analyze the function:
# ld_to_dl takes a list of dicts and transposes it to a dict of lists
# It accesses ld[0] to get keys, then iterates through ld and accesses dic[k]
# So we need a list of dicts where each dict has the same keys

ARGS = {
    "ld": [
        {"a": 1, "b": "hello", "c": torch.tensor([1.0, 2.0, 3.0, 4.0])},
        {"a": 2, "b": "world", "c": torch.tensor([5.0, 6.0, 7.0, 8.0])},
    ]
}
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 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

Also present in other papers' code

Of 3 samples lifted from this paper's linked repositories, 3 carry a function identity and 0 do not. Of the 3 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 3 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

Understanding memorisation in language models has practical and societal implications, e.g., studying models' training dynamics or preventing copyright infringements. Prior work defines memorisation as the causal effect of training with an instance on the model's ability to predict that instance. This definition relies on a counterfactual: the ability to observe what would have happened had the model not seen that instance. Existing methods struggle to provide computationally efficient and accurate estimates of this counterfactual. Further, they often estimate memorisation for a model architecture rather than for a specific model instance. This paper fills an important gap in the literature, proposing a new, principled, and efficient method to estimate memorisation based on the difference-in-differences design from econometrics. Using this method, we characterise a model's memorisation profile-its memorisation trends across training-by only observing its behaviour on a small set of instances throughout training. In experiments with the Pythia model suite, we find that memorisation (i) is stronger and more persistent in larger models, (ii) is determined by data order and learning rate, and (iii) has stable trends across model sizes, thus making memorisation in larger models predictable from smaller ones.

For agents

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

get_harvested_code_for_paper("2406.04327")
get_code_for_paper("2406.04327")
have("2406.04327")

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

curl https://syntology.ai/api/ran/2406.04327.json

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

[![Syntology run record](https://syntology.ai/api/ran/2406.04327.svg)](https://syntology.ai/paper/2406.04327)

Connect an agent — have() is free.