SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2302.02432 · ICML · 2023

Tighter Information-Theoretic Generalization Bounds from Supersamples

Ziqiao Wang, Yongyi Mao

arXiv · PDF · Open in the Atlas

Code that ran

7 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; 2 honoured the contract we drafted, 0 violated it and 5 have no contract verdict; 0 failures of our instrument; 0 no recorded run), measured 2026-11-21; 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/2302.02432.json · badge: /api/ran/2302.02432.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
hrayrhar/f-CMI canonical (classifier label)
classifier basis not recorded on the link
5 ran · 0 no recorded run
copy not recorded community
not in the paper's declared repositories
2 ran · 0 no recorded run
FunctionStatusWhere it lives
cifar_resnet50_ld_schedule Ran
ran, honoured the contract we drafted
this paper's copy was not recorded; identical code first harvested from ziqiaowanggeothe/conditional-f-information-bound
pointer only; the stored copy get_code serves, from ziqiaowanggeothe/conditional-f-information-bound, is pointer only (licence: NONE) · get_code("db64abb644176531")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

discrete_mi_est Ran
ran on a synthesized input
hrayrhar/f-CMI/modules/bound_utils.py
code served (permissive licence) · get_code("fc5d92851e4557f5")
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 float, 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 numpy as np

# discrete_mi_est expects:
# - xs: array-like of discrete values in range [0, nx)
# - ys: array-like of discrete values in range [0, ny)
# - nx: number of possible values for xs (default 2)
# - ny: number of possible values for ys (default 2)

# Create sample discrete data
np.random.seed(0)
xs = np.array([0, 1, 0, 1, 0, 1, 0, 1])
ys = np.array([0, 1, 0, 1, 1, 0, 1, 0])

ARGS = {
    "xs": xs,
    "ys": ys,
    "nx": 2,
    "ny": 2,
}
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 float, 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
estimate_fcmi_bound_classification Ran
ran on a synthesized input
hrayrhar/f-CMI/modules/bound_utils.py
code served (permissive licence) · get_code("f5fa1c6695d4e30c")
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 float64, ndim 0, 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
import numpy as np

torch.manual_seed(0)

num_examples = 3
num_classes = 2

# masks: list of arrays/tensors, one per example, each element is 0 or 1
# The function accesses masks with: ms = [p[idx] for p in masks]
# So masks is a list of sequences, and we index by example idx
masks = [
    np.array([0, 1, 0]),  # mask sequence for 3 examples
    np.array([1, 0, 1]),  # another mask sequence for 3 examples
]

# preds: list of tensors
# The function accesses preds with: ps = [p[2 * idx:2 * idx + 2] for p in preds]
# So we need tensors that can be sliced with [2*idx:2*idx+2]
# For num_examples=3, we need indices 0:2, 2:4, 4:6, so total length >= 6
# Each slice gets passed to torch.argmax(ps[i], dim=1), so shape should be (2, num_classes)
preds = [
    torch.randn(6, num_classes),  # 6 rows, 2 columns (for argmax over dim=1)
    torch.randn(6, num_classes),  # another pred tensor
]

ARGS = {
    'masks': masks,
    'preds': preds,
    'num_examples': num_examples,
    'num_classes': num_classes,
    'verbose': False,
    'return_list_of_mis': False,
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
mnist_ld_schedule Ran
ran, honoured the contract we drafted
this paper's copy was not recorded; identical code first harvested from ziqiaowanggeothe/conditional-f-information-bound
pointer only; the stored copy get_code serves, from ziqiaowanggeothe/conditional-f-information-bound, is pointer only (licence: NONE) · get_code("32ebe92e287f1e8c")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

plot_curve Ran
ran on a synthesized input
hrayrhar/f-CMI/modules/plot_utils.py
code served (permissive licence) · get_code("6f53fa4c7389e040")
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 Axes. 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
import matplotlib.pyplot as plt

# Create a mock matplotlib figure and axis
fig, ax = plt.subplots()

# Define arguments for plot_curve
xs = [1, 2, 3, 4, 5]
means = [10.0, 12.0, 11.5, 13.0, 14.5]
stds = [0.5, 0.8, 0.6, 0.7, 0.9]
marker = 'o'
label = 'test_curve'

ARGS = {
    'plt': plt,
    'ax': ax,
    'xs': xs,
    'means': means,
    'stds': stds,
    'marker': marker,
    'label': label
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
matplotlib, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Axes — a Axes output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Axes 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
plot_over_epochs Ran
ran on a synthesized input
hrayrhar/f-CMI/modules/plot_utils.py
code served (permissive licence) · get_code("20e63ae2c623c86a")
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 numpy as np
import matplotlib.pyplot as plt

# Create mock results data structure
# results[n][epoch] should be a list of dicts with 'fcmi_bound' and 'exp_gap' keys
results = {
    5: {
        1: [
            {'fcmi_bound': 0.15, 'exp_gap': 0.25},
            {'fcmi_bound': 0.18, 'exp_gap': 0.23},
            {'fcmi_bound': 0.16, 'exp_gap': 0.24},
        ],
        2: [
            {'fcmi_bound': 0.12, 'exp_gap': 0.20},
            {'fcmi_bound': 0.14, 'exp_gap': 0.19},
            {'fcmi_bound': 0.13, 'exp_gap': 0.21},
        ],
        3: [
            {'fcmi_bound': 0.10, 'exp_gap': 0.15},
            {'fcmi_bound': 0.11, 'exp_gap': 0.16},
            {'fcmi_bound': 0.09, 'exp_gap': 0.14},
        ],
    }
}

ARGS = {
    'plt': plt,
    'results': results,
    'n': 5,
    'epochs': [1, 2, 3],
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
matplotlib, numpy — 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
plot_over_n Ran
ran on a synthesized input
hrayrhar/f-CMI/modules/plot_utils.py
code served (permissive licence) · get_code("13b2515b825e9dfe")
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 numpy as np
import matplotlib.pyplot as plt

# Create mock results structure
# results[n][epoch] should be a list of dicts with 'fcmi_bound' and 'exp_gap' keys
results = {
    10: {
        0: [
            {'fcmi_bound': 0.5, 'exp_gap': 0.3},
            {'fcmi_bound': 0.6, 'exp_gap': 0.35},
            {'fcmi_bound': 0.55, 'exp_gap': 0.32},
        ]
    },
    20: {
        0: [
            {'fcmi_bound': 0.4, 'exp_gap': 0.25},
            {'fcmi_bound': 0.45, 'exp_gap': 0.28},
            {'fcmi_bound': 0.42, 'exp_gap': 0.26},
        ]
    },
    30: {
        0: [
            {'fcmi_bound': 0.3, 'exp_gap': 0.2},
            {'fcmi_bound': 0.35, 'exp_gap': 0.22},
            {'fcmi_bound': 0.32, 'exp_gap': 0.21},
        ]
    }
}

# Create matplotlib objects
fig_dummy, ax_dummy = plt.subplots()

ARGS = {
    'plt': plt,
    'results': results,
    'ns': [10, 20, 30],
    'epoch': 0,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
matplotlib, numpy — 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

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, 2 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 2 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.

FunctionAlso present inEarliest-datedOthers named
cifar_resnet50_ld_schedule 1 other paper
1 later
2410.22887 (2024-10)
Generalization Bounds via Conditional f -Information
mnist_ld_schedule 1 other paper
1 later
2410.22887 (2024-10)
Generalization Bounds via Conditional f -Information

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 2026-11-21.

Repositories linked to this paper

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

Abstract

In this work, we present a variety of novel information-theoretic generalization bounds for learning algorithms, from the supersample setting of Steinke & Zakynthinou (2020)-the setting of the "conditional mutual information" framework. Our development exploits projecting the loss pair (obtained from a training instance and a testing instance) down to a single number and correlating loss values with a Rademacher sequence (and its shifted variants). The presented bounds include square-root bounds, fastrate bounds, including those based on variance and sharpness, and bounds for interpolating algorithms etc. We show theoretically or empirically that these bounds are tighter than all informationtheoretic bounds known to date on the same supersample setting.

For agents

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

get_harvested_code_for_paper("2302.02432")
get_code_for_paper("2302.02432")
have("2302.02432")

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

curl https://syntology.ai/api/ran/2302.02432.json

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

[![Syntology run record](https://syntology.ai/api/ran/2302.02432.svg)](https://syntology.ai/paper/2302.02432)

Connect an agent — have() is free.