SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2407.10964 · NeurIPS · 2024

No Train, all Gain: Self-Supervised Gradients Improve Deep Frozen Representations

Yuki Asano, Andrei Bursuc, Spyros Gidaris, Walter Simoncini

arXiv · PDF · Open in the Atlas

Code that ran

7 of 12 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, 1 violated it and 0 have no contract verdict; 4 failures of our instrument; 5 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/2407.10964.json · badge: /api/ran/2407.10964.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
WalterSimoncini/fungivision canonical (classifier label)
classifier basis: census
5 ran · 2 no recorded run
waltersimoncini/fungivision canonical (classifier label)
classifier basis: classifier_2025
2 ran · 3 no recorded run
FunctionStatusWhere it lives
_layer_type Ran
ran; our contract draft was wrong, not the code
waltersimoncini/fungivision/fungivision/gradients/base_extractor.py
code served (permissive licence) · get_code("104dacd44ae83925")
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 str. 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 nn.Module instance
layer = nn.Linear(4, 3)

ARGS = {
    "layer": layer
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
generate_projection_matrix Ran
ran; our contract draft was wrong, not the code
WalterSimoncini/fungivision/fungivision/utils/compression.py
code served (permissive licence) · get_code("1f2b26d7aa212e61")
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 Tensor, ndim 2, dtype family float, shape (4, 3) 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
from typing import Tuple

torch.manual_seed(0)

ARGS = {
    "dims": (4, 3),
    "device": torch.device('cpu')
}
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 Tensor, ndim 2, dtype family float, shape (4, 3) — shape (4, 3) 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
is_supported Ran
ran, violated the contract we drafted
WalterSimoncini/fungivision/fungivision/utils/autograd_hacks.py
code served (permissive licence) · get_code("01e52b8e916acca2")
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 bool, 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 torch.nn as nn

torch.manual_seed(0)

# Test with a Linear layer (supported)
linear_layer = nn.Linear(4, 3)

# Test with a Conv2d layer (supported)
conv_layer = nn.Conv2d(3, 16, kernel_size=3)

# Test with a MultiheadAttention layer (supported)
mha_layer = nn.MultiheadAttention(embed_dim=8, num_heads=2, batch_first=True)

# Test with an unsupported layer (e.g., ReLU)
unsupported_layer = nn.ReLU()

ARGS = {
    "layer": linear_layer
}
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 bool, 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
model_feature_dim Ran
ran, honoured the contract we drafted
WalterSimoncini/fungivision/fungivision/utils/misc.py
code served (permissive licence) · get_code("3db9ee14869a6845")
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 int, ndim 0, dtype family int, 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 torch.nn as nn

torch.manual_seed(0)

# Create a minimal feature extractor model
class SimpleFeatureExtractor(nn.Module):
    def __init__(self, output_dim=64):
        super().__init__()
        self.conv1 = nn.Conv2d(3, 16, kernel_size=3, padding=1)
        self.pool = nn.AdaptiveAvgPool2d((1, 1))
        self.fc = nn.Linear(16, output_dim)
    
    def forward(self, x):
        x = self.conv1(x)
        x = self.pool(x)
        x = x.view(x.size(0), -1)
        x = self.fc(x)
        return x

model = SimpleFeatureExtractor(output_dim=64)
device = torch.device('cpu')
image_size = 224

ARGS = {
    'model': model,
    'device': device,
    'image_size': image_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 int, ndim 0, dtype family int, 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
preprocess_layer_paths Ran
ran; our contract draft was wrong, not the code
WalterSimoncini/fungivision/fungivision/utils/autograd_hacks.py
code served (permissive licence) · get_code("30d027ce5196ab7c")
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 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 torch.nn as nn

torch.manual_seed(0)

# The function takes a list of layer paths (strings) and processes them
# It checks if 'self_attention' is in the path and adds extra layers as needed
# No model or complex objects are actually used - just strings

ARGS = {
    "paths": [
        "layer.0.self_attention.out_proj",
        "layer.1.mlp",
        "layer.2.self_attention.qkv",
    ]
}
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 list, ndim 1, dtype family float — 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
rgetattr Ran
ran; our contract draft was wrong, not the code
waltersimoncini/fungivision/fungivision/gradients/base_extractor.py
code served (permissive licence) · get_code("7adb31904977e81b")
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 Linear. 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
import functools
from typing import List, Any

torch.manual_seed(0)

# Create a nested module structure to test rgetattr
class SimpleBlock(nn.Module):
    def __init__(self):
        super().__init__()
        self.proj = nn.Linear(4, 3)
    
class SimpleAttention(nn.Module):
    def __init__(self):
        super().__init__()
        self.attn = SimpleBlock()

class SimpleEncoder(nn.Module):
    def __init__(self):
        super().__init__()
        self.blocks = nn.ModuleList([SimpleAttention() for _ in range(2)])

class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.encoder = SimpleEncoder()

# Create the model
model = SimpleModel()

# Test case 1: Access nested attribute with dot-separated path
obj = model
attr = "encoder.blocks.0.attn.proj"

ARGS = {
    "obj": obj,
    "attr": attr
}
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 Linear — a Linear output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Linear 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
suggested_scaling_factor Ran
ran, honoured the contract we drafted
WalterSimoncini/fungivision/fungivision/utils/compression.py
code served (permissive licence) · get_code("ba88ce6fa6ae61d3")
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 Tensor, 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

torch.manual_seed(0)

ARGS = {
    "projection_dim": 16
}
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 Tensor, 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
BaseGradientExtractor Not yet run
unverified (harvested, no recorded run)
waltersimoncini/fungivision/fungivision/gradients/base_extractor.py
code served (permissive licence) · get_code("8c517a7e744ed152")
How it ran

no invocation recorded.

clear_backprops Not yet run
unverified (harvested, no recorded run)
waltersimoncini/fungivision/fungivision/gradients/base_extractor.py
code served (permissive licence) · get_code("30dc14b4b04ac056")
How it ran

no invocation recorded.

compute_grad1 Not yet run
unverified (harvested, no recorded run)
waltersimoncini/fungivision/fungivision/gradients/base_extractor.py
code served (permissive licence) · get_code("546242bc466400c2")
How it ran

no invocation recorded.

rsetattr Not yet run
unverified (harvested, no recorded run)
WalterSimoncini/fungivision/fungivision/utils/misc.py
code served (permissive licence) · get_code("4632ca8fbf0dd221")
How it ran

no invocation recorded.

symsqrt Not yet run
unverified (harvested, no recorded run)
WalterSimoncini/fungivision/fungivision/utils/autograd_hacks.py
code served (permissive licence) · get_code("56afcb3496c8fd91")
How it ran

no invocation recorded.

Also present in other papers' code

Of 12 samples lifted from this paper's linked repositories, 12 carry a function identity and 0 do not. Of the 12 distinct functions with an identity, 2 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 0 later) and 10 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
is_supported 3 other papers
3 dated earlier
2205.14495 (2022-05)
Task-Agnostic Continual Reinforcement Learning: Gaining Insights and Overcoming Challenges
2206.02927 (2022-06)
Spectral Bias Outside the Training Set for Deep Networks in the Kernel Regime
2403.10318 (2024-03)
pTNAS: Progressive Neural Architecture Search for Tabular Data
symsqrt 3 other papers
3 dated earlier
2205.14495 (2022-05)
Task-Agnostic Continual Reinforcement Learning: Gaining Insights and Overcoming Challenges
2206.02927 (2022-06)
Spectral Bias Outside the Training Set for Deep Networks in the Kernel Regime
2403.10318 (2024-03)
pTNAS: Progressive Neural Architecture Search for Tabular Data

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 introduces FUNGI, Features from UNsupervised GradIents, a method to enhance the features of transformer encoders by leveraging self-supervised gradients. Our method is simple: given any pretrained model, we first compute gradients from various self-supervised objectives for each input. These gradients are projected to a lower dimension and then concatenated with the model's output embedding. The resulting features are evaluated on k-nearest neighbor classification over 11 datasets from vision, 5 from natural language processing, and 2 from audio. Across backbones spanning various sizes and pretraining strategies, FUNGI features provide consistent performance improvements over the embeddings. We also show that using FUNGI features can benefit linear classification, clustering and image retrieval, and that they significantly improve the retrieval-based in-context scene understanding abilities of pretrained models, for example improving upon DINO by +17% for semantic segmentation -without any training. Code is available at https://github.com/WalterSimoncini/fungivision.

For agents

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

get_harvested_code_for_paper("2407.10964")
get_code_for_paper("2407.10964")
have("2407.10964")

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

curl https://syntology.ai/api/ran/2407.10964.json

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

[![Syntology run record](https://syntology.ai/api/ran/2407.10964.svg)](https://syntology.ai/paper/2407.10964)

Connect an agent — have() is free.