SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2603.29183 · ICML · 2026

IMPACT: Influence Modeling for Open-Set Time Series Anomaly Detection

Hongzuo Xu, Yijie Wang, Guansong Pang, Xiaoli Li, Weixuan Liang, Xiaohui Zhou

arXiv · PDF · Open in the Atlas

Code that ran

9 of 10 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (6 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 7 have no contract verdict; 2 failures of our instrument; 1 no recorded run), measured 2026-11-22; 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/2603.29183.json · badge: /api/ran/2603.29183.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
mala-lab/IMPACT community
Papers with Code archive
9 ran · 1 no recorded run
FunctionStatusWhere it lives
Chomp1d Ran
ran on a synthesized input · constructed an object; no result was computed
mala-lab/IMPACT/model/impact.py
code served (permissive licence) · get_code("0e741ef5f756b23a")
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, Chomp1d: 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)

# Chomp1d takes a chomp_size parameter
# The forward method expects input x with shape (batch, channels, length)
# and removes chomp_size elements from the end of the length dimension

# Create a small test tensor: batch=2, channels=4, length=10
test_input = torch.randn(2, 4, 10)

# chomp_size should be less than the sequence length
chomp_size = 3

ARGS = {
    'chomp_size': chomp_size
}
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 Chomp1d — 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 Chomp1d 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
DSADLoss Ran
ran on a synthesized input · constructed an object; no result was computed
mala-lab/IMPACT/model/impact.py
code served (permissive licence) · get_code("b9b11200930d3efa")
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, DSADLoss: 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)

# The DSADLoss expects:
# - c: center point tensor (used in forward as self.c)
# - eta: float hyperparameter
# - eps: float for numerical stability
# - reduction: string for reduction mode
# - rep: representation tensor in forward (batch_size, dim)
# - semi_targets: optional tensor of same batch size as rep

# Create a small center tensor (dimension 4)
c = torch.randn(4)

# Create representation tensor (batch_size=2, dim=4)
rep = torch.randn(2, 4)

# Create semi_targets tensor (batch_size=2)
# Contains binary values (0 or 1) for semi-supervised learning
semi_targets = torch.tensor([0, 1], dtype=torch.float32)

ARGS = {
    'c': c,
    'eta': 1.0,
    'eps': 1e-6,
    'reduction': 'mean'
}
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 DSADLoss — 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 DSADLoss 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
HolisticHead Ran
ran on a synthesized input · constructed an object; no result was computed
mala-lab/IMPACT/model/impact.py
code served (permissive licence) · get_code("483a06f3a704b8d8")
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, HolisticHead: 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)

ARGS = {
    'in_dim': 4,
    'score_dim': 3,
    'dropout': 0.5
}
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 HolisticHead — 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 HolisticHead 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
IMPACTNet Ran
ran on a synthesized input · constructed an object; no result was computed
mala-lab/IMPACT/model/impact.py
code served (permissive licence) · get_code("a23e76ec9ac6e56e")
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, IMPACTNet: 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)

ARGS = {
    'n_features': 4,
    'n_hidden': 64,
    'n_output': 64,
    'score_dim': 3,
    'kernel_size': 2,
    'bias': True,
    'maxpool_out_channels': 8,
    'normalize_embedding': True,
    'ref_size': 5,
    'dropout': 0.2,
    'activation': 'ReLU'
}
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 IMPACTNet — 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 IMPACTNet 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
Multi_DeviationLoss Ran
ran on a synthesized input
mala-lab/IMPACT/model/impact.py
code served (permissive licence) · get_code("beb196a3abce538a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 0, dtype family float; no output shape is stated. 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 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 torch.nn as nn

class Multi_DeviationLoss(nn.Module):

    def __init__(self, score_dim):
        super().__init__()
        self.score_dim = score_dim

    def forward(self, y_pred, y_true, confidence_margin=5.0):
        mean = torch.tensor([0.0] * self.score_dim).cuda()
        std = torch.tensor([1.0] * self.score_dim).cuda()
        y_true = y_true.unsqueeze(1).repeat(1, self.score_dim)
        ref = torch.normal(mean=mean.expand(5000, -1), std=std.expand(5000, -1))
        dev = (y_pred - torch.mean(ref, dim=0)) / torch.std(ref, dim=0)
        inlier_loss = torch.abs(dev)
        outlier_loss = torch.abs((confidence_margin - dev).clamp_(min=0.0))
        dev_loss = (1 - y_true) * inlier_loss + y_true * outlier_loss
        return torch.mean(dev_loss)


INIT_ARGS = {
    'score_dim': 4
}

ARGS = {
    'y_pred': torch.randn(2, 4),
    'y_true': torch.tensor([0, 1], dtype=torch.float32),
    'confidence_margin': 5.0
}

if __name__ == '__main__':
    model = Multi_DeviationLoss(**INIT_ARGS)
    try:
        output = model(**ARGS)
        print(f"Output: {output}")
    except Exception as e:
        print(f"Expected error (CUDA not available): {type(e).__name__}: {e}")
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 0, dtype family float — no output shape is stated: observe_signature.py observed this class with its constructor sizes canonicalised where it could, which the node does not record, so its shape is not stated as this setup block's
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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
TCNEncoder Ran
ran on a synthesized input · constructed an object; no result was computed
mala-lab/IMPACT/model/impact.py
code served (permissive licence) · get_code("fd576b23b5d20b65")
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, TCNEncoder: 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)

ARGS = {
    'n_features': 4,
    'n_hidden': '500,100',
    'n_output': 20,
    'kernel_size': 2,
    'bias': False,
    'dropout': 0.2,
    'activation': 'ReLU',
    'maxpool_out_channels': 1,
    'normalize_embedding': True,
}
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 TCNEncoder — 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 TCNEncoder 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
TcnResidualBlock Ran
ran on a synthesized input · constructed an object; no result was computed
mala-lab/IMPACT/model/impact.py
code served (permissive licence) · get_code("36371c7e0899a9ca")
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, TcnResidualBlock: 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)

ARGS = {
    'n_inputs': 4,
    'n_outputs': 8,
    'kernel_size': 3,
    'stride': 1,
    'dilation': 1,
    'padding': 1,
    'dropout': 0.2,
    'activation': 'ReLU',
    'bias': True
}
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 TcnResidualBlock — 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 TcnResidualBlock 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
_instantiate_class Ran
ran; our contract draft was wrong, not the code
mala-lab/IMPACT/model/impact.py
code served (permissive licence) · get_code("a2bfe81fe3467af3")
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 OrderedDict. 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 importlib

# The function _instantiate_class uses:
# - module_name: a string naming a module to import
# - class_name: a string naming a class within that module
# It then imports the module, gets the class, and instantiates it with no arguments

# We need to provide a module and class that exist and can be instantiated with no args

# Using a built-in module and class that's always available
ARGS = {
    "module_name": "collections",
    "class_name": "OrderedDict"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type OrderedDict — a OrderedDict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a OrderedDict 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
hvp Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
mala-lab/IMPACT/model/impact.py
code served (permissive licence) · get_code("de0a2f38dd099d3b")
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. 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
from torch.autograd import grad

torch.manual_seed(0)

# Create a simple computational graph where y depends on w
# y = sum(w^2) is a simple scalar output
w = [
    torch.tensor([1.0, 2.0, 3.0, 4.0], requires_grad=True),
    torch.tensor([5.0, 6.0, 7.0, 8.0], requires_grad=True)
]

# Compute y as a scalar that depends on w
y = torch.sum(w[0]**2) + torch.sum(w[1]**2)

# Create v vectors with the same shapes as w
v = [
    torch.tensor([0.1, 0.2, 0.3, 0.4]),
    torch.tensor([0.5, 0.6, 0.7, 0.8])
]

ARGS = {
    'y': y,
    'w': w,
    'v': v
}
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 — a tuple output has no array shape of its own; none is 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
IMPACT Not yet run
unverified (harvested, no recorded run)
mala-lab/IMPACT/model/impact.py
code served (permissive licence) · get_code("d4ffe5c1b46b92a4")
How it ran

no invocation recorded.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 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 9 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
_instantiate_class
brick of this paper's assembly: helper
2 other papers
2 dated earlier · 2 only as an assembly brick
2005.02359 (2020-05)
Classification-Based Anomaly Detection for General Data
2305.16114 (2023-05)
Fascinating Supervisory Signals and Where to Find Them: Deep Anomaly Detection with Scale Learning

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

Open-set anomaly detection (OSAD) is an emerging paradigm designed to utilize limited labeled data from anomaly classes seen in training to identify both seen and unseen anomalies during testing. Current approaches rely on simple augmentation methods to generate pseudo anomalies that replicate unseen anomalies. Despite being promising in image data, these methods are found to be ineffective in time series data due to the failure to preserve its sequential nature, resulting in trivial or unrealistic anomaly patterns. They are further plagued when the training data is contaminated with unlabeled anomalies. This work introduces IMPACT, a novel framework that leverages influence modeling for open-set time series anomaly detection, to tackle these challenges. The key insight is to i) learn an influence function that can accurately estimate the impact of individual training samples on the modeling, and then ii) leverage these influence scores to generate semantically divergent yet realistic unseen anomalies for time series while repurposing high-influential samples as supervised anomalies for anomaly decontamination. Extensive experiments show that IMPACT significantly outperforms existing state-of-the-art methods, showing superior accuracy under varying OSAD settings and contamination rates. Code is available at https://github.com/mala-lab/IMPACT.

For agents

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

get_harvested_code_for_paper("2603.29183")
get_code_for_paper("2603.29183")
have("2603.29183")

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

curl https://syntology.ai/api/ran/2603.29183.json

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

[![Syntology run record](https://syntology.ai/api/ran/2603.29183.svg)](https://syntology.ai/paper/2603.29183)

Connect an agent — have() is free.