SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2202.10054 · ICLR · 2022

Fine-Tuning can Distort Pretrained Features and Underperform Out-of-Distribution

Aditi Raghunathan, Percy Liang, Tengyu Ma, Ananya Kumar, Robbie Jones

arXiv · PDF · Open in the Atlas

Code that ran

1 of 4 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; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 1 failure of our instrument; 3 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/2202.10054.json · badge: /api/ran/2202.10054.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
hsh1108/daft community
Papers with Code archive
1 ran · 0 no recorded run
skhnha/daft community
Papers with Code archive
0 ran · 3 no recorded run
FunctionStatusWhere it lives
train_epoch Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
hsh1108/daft/lp_sgd.py
code served (permissive licence) · get_code("5ceddfee16e557e7")
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 dict. 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 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 logging

torch.manual_seed(0)

# Create a mock network with get_feature_extractor() and get_last_layer() methods
class MockFeatureExtractor(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 8)
    
    def forward(self, x):
        return self.linear(x)

class MockLastLayer(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(8, 3)
    
    def forward(self, x):
        return self.linear(x)

class MockNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.feature_extractor = MockFeatureExtractor()
        self.last_layer = MockLastLayer()
    
    def get_feature_extractor(self):
        return self.feature_extractor
    
    def get_last_layer(self):
        return self.last_layer
    
    def forward(self, x):
        x = self.feature_extractor(x)
        x = self.last_layer(x)
        return x

# Create config dict
config = {
    'log_interval': 2
}

# Create results dict
results = {
    'train_acc': [],
    'train_loss': []
}

# Create network
net = MockNet()

# Create train_loader as list of (image, label) tuples
train_loader = [
    (torch.randn(2, 4), torch.randint(0, 3, (2,))),
    (torch.randn(2, 4), torch.randint(0, 3, (2,))),
    (torch.randn(2, 4), torch.randint(0, 3, (2,)))
]

# Create optimizer
optimizer = torch.optim.SGD(net.parameters(), lr=0.01)

# Create criterion
criterion = nn.CrossEntropyLoss()

ARGS = {
    'config': config,
    'results': results,
    'net': net,
    'train_loader': train_loader,
    'optimizer': optimizer,
    'criterion': criterion
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
main Not yet run
unverified (harvested, no recorded run)
skhnha/daft/ft_sgd.py
code served (permissive licence) · get_code("0d3bf4b33434cf07")
How it ran

no invocation recorded.

test_epoch Not yet run
unverified (harvested, no recorded run)
skhnha/daft/ft_sgd.py
code served (permissive licence) · get_code("78430fe0e766f96a")
How it ran

no invocation recorded.

train_epoch Not yet run
unverified (harvested, no recorded run)
skhnha/daft/ft_sgd.py
code served (permissive licence) · get_code("df6396fcc8d20e79")
How it ran

no invocation recorded.

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, 0 are also present in other papers whose code we hold (0 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.

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

When transferring a pretrained model to a downstream task, two popular methods are full fine-tuning (updating all the model parameters) and linear probing (updating only the last linear layer-the "head"). It is well known that fine-tuning leads to better accuracy in-distribution (ID). However, in this paper, we find that fine-tuning can achieve worse accuracy than linear probing out-of-distribution (OOD) when the pretrained features are good and the distribution shift is large. On 10 distribution shift datasets (Breeds-Living17, Breeds-Entity30, DomainNet, CIFAR → STL, CIFAR10.1, FMoW, ImageNetV2, ImageNet-R, ImageNet-A, ImageNet-Sketch), fine-tuning obtains on average 2% higher accuracy ID but 7% lower accuracy OOD than linear probing. We show theoretically that this tradeoff between ID and OOD accuracy arises even in a simple setting: fine-tuning overparameterized two-layer linear networks. We prove that the OOD error of fine-tuning is high when we initialize with a fixed or random head-this is because while fine-tuning learns the head, the lower layers of the neural network change simultaneously and distort the pretrained features. Our analysis suggests that the easy two-step strategy of linear probing then full fine-tuning (LP-FT), sometimes used as a fine-tuning heuristic, combines the benefits of both fine-tuning and linear probing. Empirically, LP-FT outperforms both fine-tuning and linear probing on the above datasets (1% better ID, 10% better OOD than full fine-tuning).

For agents

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

get_harvested_code_for_paper("2202.10054")
get_code_for_paper("2202.10054")
have("2202.10054")

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

curl https://syntology.ai/api/ran/2202.10054.json

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

[![Syntology run record](https://syntology.ai/api/ran/2202.10054.svg)](https://syntology.ai/paper/2202.10054)

Connect an agent — have() is free.