SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2409.09778 · NeurIPS · 2025

Rewind-to-Delete: Certified Machine Unlearning for Nonconvex Functions

Diego Klabjan, Siqiao Mu

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; 2 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 2 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/2409.09778.json · badge: /api/ran/2409.09778.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
siqiaomu/r2d canonical (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
copy not recorded community
not in the paper's declared repositories
1 ran · 0 no recorded run
zhangbinchi/certified-deep-unlearning unrelated (classifier label)
classifier basis: classifier_2025
0 ran · 1 no recorded run
FunctionStatusWhere it lives
compute_test_error Ran
ran, honoured the contract we drafted
siqiaomu/r2d/r2d.py
code served (permissive licence) · get_code("6c94779e31eb4e12")
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 float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 torch.nn as nn

torch.manual_seed(0)

# Build a minimal model that:
# - Takes input data
# - Returns output with shape (batch_size, num_classes) where num_classes > 1
# - Has eval() and no_grad() compatibility
class MinimalModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)  # 4 input features, 3 classes
    
    def forward(self, x):
        return self.linear(x)

# Build a minimal data loader
# Each item is (data, target, identity)
# data: (batch_size, 4) - 4 features
# target: (batch_size,) - class labels
# identity: arbitrary (not used in function)
def create_loader():
    loader_data = []
    for i in range(2):  # 2 batches
        data = torch.randn(2, 4)  # batch_size=2, features=4
        target = torch.randint(0, 3, (2,))  # batch_size=2, 3 classes
        identity = torch.tensor([i, i])  # dummy identity
        loader_data.append((data, target, identity))
    return loader_data

# Create loader with dataset attribute
class SimpleLoader:
    def __init__(self, data):
        self.data = data
        self.dataset = type('obj', (object,), {'__len__': lambda self: sum(len(batch[0]) for batch in data)})()
    
    def __iter__(self):
        return iter(self.data)
    
    def __len__(self):
        return len(self.data)

loader = SimpleLoader(create_loader())
model = MinimalModel()
device = torch.device('cpu')

ARGS = {
    'model': model,
    'loader': loader,
    'device': device
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, scipy, sklearn, torch — 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
get_error Ran
ran; our contract draft was wrong, not the code
siqiaomu/r2d/r2d.py
code served (permissive licence) · get_code("de35b5dbce9d27e9")
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 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
import numpy as np

torch.manual_seed(0)

# The function expects:
# - output: a tensor with shape (batch_size, num_classes) where num_classes > 1
# - target: a tensor with shape (batch_size,) or (batch_size, 1) containing class indices

# Create output tensor with shape (2, 4) - batch size 2, 4 classes
output = torch.randn(2, 4)

# Create target tensor with shape (2,) - batch size 2, class indices
target = torch.tensor([0, 2])

ARGS = {
    "output": output,
    "target": target,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, scipy, sklearn, 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
h_function Ran
ran, honoured the contract we drafted
siqiaomu/r2d/r2d.py
code served (permissive licence) · get_code("8b0a5db3b9d9ea07")
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 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 torch
import math
import numpy as np
from math import exp, sqrt
from scipy.special import erf, log_ndtr
from sklearn.metrics import roc_auc_score

torch.manual_seed(0)

# h_function takes 6 scalar parameters: K, eta, L, m, n, T
# All appear to be numeric values used in arithmetic operations

ARGS = {
    "K": 2,           # integer, exponent
    "eta": 0.1,       # float, learning rate-like parameter
    "L": 1.5,         # float, Lipschitz constant-like parameter
    "m": 3,           # integer, count parameter
    "n": 10,          # integer, count parameter (must be > m)
    "T": 5,           # integer, time/iteration parameter (must be >= K)
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, scipy, sklearn, torch — 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
hvp Ran
ran; our contract draft was wrong, not the code
this paper's copy was not recorded; identical code first harvested from zhangbinchi/certified-deep-unlearning
pointer only · get_code("e261328dc1f89c4a")
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.

newton_update Not yet run
unverified (harvested, no recorded run)
zhangbinchi/certified-deep-unlearning/unlearn.py
code served (permissive licence) · get_code("1acec0665cb7fb8e")
How it ran

no invocation recorded.

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, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 0 later) and 4 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
hvp
brick of this paper's assembly: helper
1 other paper
1 dated earlier
2408.00920 (2024-08)
Towards Certified Unlearning for Deep Neural Networks

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

Machine unlearning algorithms aim to efficiently remove data from a model without retraining it from scratch, in order to remove corrupted or outdated data or respect a user's "right to be forgotten." Certified machine unlearning is a strong theoretical guarantee based on differential privacy that quantifies the extent to which an algorithm erases data from the model weights. In contrast to existing works in certified unlearning for convex or strongly convex loss functions, or nonconvex objectives with limiting assumptions, we propose the first, first-order, black-box (i.e., can be applied to models pretrained with vanilla gradient descent) algorithm for unlearning on general nonconvex loss functions, which unlearns by "rewinding" to an earlier step during the learning process before performing gradient descent on the loss function of the retained data points. We prove (ϵ, δ) certified unlearning and performance guarantees that establish the privacy-utility-complexity tradeoff of our algorithm, and we prove generalization guarantees for functions that satisfy the Polyak-Lojasiewicz inequality. Finally, we demonstrate the superior performance of our algorithm compared to existing methods, within a new experimental framework that more accurately reflects unlearning user data in practice.

For agents

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

get_harvested_code_for_paper("2409.09778")
get_code_for_paper("2409.09778")
have("2409.09778")

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

curl https://syntology.ai/api/ran/2409.09778.json

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

[![Syntology run record](https://syntology.ai/api/ran/2409.09778.svg)](https://syntology.ai/paper/2409.09778)

Connect an agent — have() is free.