SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2407.12993 · ICML · 2024

Improving SAM Requires Rethinking its Optimization Formulation

Fabian Latorre, Volkan Cevher, Thomas Pethick, Kimon Antonakopoulos, Wanyun Xie

arXiv · PDF · Open in the Atlas

Code that ran

4 of 4 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (1 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 1 has no contract verdict; 3 failures 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/2407.12993.json · badge: /api/ran/2407.12993.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
lions-epfl/bisam canonical (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
LIONS-EPFL/BiSAM marked official in Papers with Code
Papers with Code archive
1 ran · 0 no recorded run
FunctionStatusWhere it lives
SAM Ran
ran on a synthesized input · constructed an object; no result was computed
LIONS-EPFL/BiSAM/utility/sam.py
code served (permissive licence) · get_code("aacd1c60e21babdd")
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 an instance of its own class, SAM: the run constructed the object and did not call it, so no result was computed. 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 torch.nn as nn

torch.manual_seed(0)

# Create a simple model with parameters
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear1 = nn.Linear(4, 3)
        self.linear2 = nn.Linear(3, 2)
    
    def forward(self, x):
        x = torch.relu(self.linear1(x))
        x = self.linear2(x)
        return x

# Instantiate model
model = SimpleModel()

# Create parameters list
params = list(model.parameters())

# Base optimizer class (SGD)
base_optimizer = torch.optim.SGD

# kwargs for base optimizer
optimizer_kwargs = {
    'lr': 0.01,
    'momentum': 0.9
}

ARGS = {
    'params': params,
    'base_optimizer': base_optimizer,
    'rho': 0.05,
    'adaptive': False,
    'lr': 0.01,
    'momentum': 0.9
}
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 SAM — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a SAM 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
perturbation_loss_log Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
lions-epfl/bisam/loss_fnc.py
code served (permissive licence) · get_code("83634a07e94d9e91")
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 Tensor, ndim 1, dtype family float, shape (2,) 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 torch.nn as nn
import torch.nn.functional as F
import math

torch.manual_seed(0)

# Create pred: a tensor of shape (batch_size, num_classes)
# The function indexes pred[i][targets[i]], so we need at least 2D
pred = torch.randn(2, 4)

# Create targets: a list of indices, one per batch element
# Each index should be valid (0 to num_classes-1)
targets = [1, 2]

# mu parameter
mu = 1

ARGS = {
    "pred": pred,
    "targets": targets,
    "mu": mu
}
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 Tensor, ndim 1, dtype family float, shape (2,) — shape (2,) 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
perturbation_loss_tanh Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
lions-epfl/bisam/loss_fnc.py
code served (permissive licence) · get_code("dd73f8d81a55afb2")
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 Tensor, ndim 1, dtype family float, shape (2,) 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 torch.nn as nn

torch.manual_seed(0)

# Analyze the function:
# - pred: indexed with pred[i], then pred[i][targets[i]], then used in torch.tanh and torch.logsumexp
#   This suggests pred is a 2D tensor (batch_size x num_classes)
# - targets: indexed with targets[i], values used as indices into pred[i]
#   This suggests targets is a 1D tensor of class indices
# - mu: scalar, multiplied with tanh result
# - alpha: scalar, multiplied with pred before tanh

# Create small tensors
batch_size = 2
num_classes = 4

# pred: shape (batch_size, num_classes)
pred = torch.randn(batch_size, num_classes)

# targets: shape (batch_size,) with valid class indices
targets = torch.tensor([0, 2])

# Scalars
mu = 10
alpha = 0.1

ARGS = {
    "pred": pred,
    "targets": targets,
    "mu": mu,
    "alpha": alpha
}
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 Tensor, ndim 1, dtype family float, shape (2,) — shape (2,) 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
smooth_crossentropy Ran
ran; our contract draft was wrong, not the code
lions-epfl/bisam/loss_fnc.py
code served (permissive licence) · get_code("e34a3257087e025a")
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 Tensor, ndim 1, dtype family float, shape (2,) 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 torch.nn as nn
import torch.nn.functional as F

torch.manual_seed(0)

# Create pred: logits tensor of shape (batch_size=2, n_class=4)
pred = torch.randn(2, 4)

# Create gold: class indices tensor of shape (batch_size=2,)
# Values should be in range [0, n_class-1]
gold = torch.tensor([0, 2])

# smoothing parameter
smoothing = 0.1

ARGS = {
    "pred": pred,
    "gold": gold,
    "smoothing": smoothing
}
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 Tensor, ndim 1, dtype family float, shape (2,) — shape (2,) 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

Also present in other papers' code

Of 4 samples lifted from this paper's linked repositories, 4 carry a function identity and 0 do not. Of the 4 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 2 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
perturbation_loss_tanh 1 other paper
1 later
2410.10683 (2024-10)
SAMPa: Sharpness-aware Minimization Parallelized
smooth_crossentropy 1 other paper
1 later
2410.10683 (2024-10)
SAMPa: Sharpness-aware Minimization Parallelized

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

This paper rethinks Sharpness-Aware Minimization (SAM), which is originally formulated as a zero-sum game where the weights of a network and a bounded perturbation try to minimize/maximize, respectively, the same differentiable loss. To fundamentally improve this design, we argue that SAM should instead be reformulated using the 0-1 loss. As a continuous relaxation, we follow the simple conventional approach where the minimizing (maximizing) player uses an upper bound (lower bound) surrogate to the 0-1 loss. This leads to a novel formulation of SAM as a bilevel optimization problem, dubbed as BiSAM. BiSAM with newly designed lower-bound surrogate loss indeed constructs stronger perturbation. Through numerical evidence, we show that BiSAM consistently results in improved performance when compared to the original SAM and variants, while enjoying similar computational complexity. Our code is available at https://github.com/ LIONS-EPFL/BiSAM.

For agents

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

get_harvested_code_for_paper("2407.12993")
get_code_for_paper("2407.12993")
have("2407.12993")

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

curl https://syntology.ai/api/ran/2407.12993.json

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

[![Syntology run record](https://syntology.ai/api/ran/2407.12993.svg)](https://syntology.ai/paper/2407.12993)

Connect an agent — have() is free.