SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2310.19152 · NeurIPS · 2023

BERT Lost Patience Won't Be Robust to Adversarial Slowdown

Sanghyun Hong, Zachary Coalson, Gabriel Ritter, Rakesh Bobba

arXiv · PDF · Open in the Atlas

Code that ran

3 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; 0 honoured the contract we drafted, 1 violated it and 0 have no contract verdict; 2 failures of our instrument; 2 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/2310.19152.json · badge: /api/ran/2310.19152.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
ztcoalson/waffle canonical (classifier label)
classifier basis: census
3 ran · 2 no recorded run
FunctionStatusWhere it lives
avg_exit Ran
ran, violated the contract we drafted
ztcoalson/waffle/project/attack_exit.py
code served (permissive licence) · get_code("839b9d0ef5c35d5c")
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 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

torch.manual_seed(0)

# Construct valid arguments for avg_exit function
# The function expects 'stats' - a list-like object where:
# - It iterates through indices (lidx) from 0 to len(stats)-1
# - It accesses stats[lidx] as a numeric value
# - It sums all values in stats
# - It performs arithmetic operations on these values

stats = [0.5, 1.2, 0.8, 0.3]

ARGS = {
    "stats": stats
}
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 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
entropy Ran
ran; our contract draft was wrong, not the code
ztcoalson/waffle/project/deebert/modeling_highway_bert.py
code served (permissive licence) · get_code("63301b5cf89d7b03")
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 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

torch.manual_seed(0)

# Create a small batch of logits
# Shape: (batch_size=2, num_classes=4)
x = torch.randn(2, 4)

ARGS = {
    "x": x
}
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 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
get_tasks Ran
ran; our contract draft was wrong, not the code
ztcoalson/waffle/download_glue_data.py
pointer only (licence: MIT); the stored copy get_code serves, from nyu-mll/GLUE-baselines, is pointer only (licence: NONE) · get_code("3caf26b311c98658")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from nyu-mll/GLUE-baselines, and GitHub detected no licence in that repository when it was checked. 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.

efficacy Not yet run
unverified (harvested, no recorded run)
ztcoalson/waffle/project/attack_exit.py
code served (permissive licence) · get_code("49a7658246241580")
How it ran

no invocation recorded.

layer_info Not yet run
unverified (harvested, no recorded run)
ztcoalson/waffle/project/attack_univ_exit.py
code served (permissive licence) · get_code("a9744f188d4721d6")
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, 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 3 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
get_tasks 13 other papers
12 dated earlier, 1 same year, order not determinable
1711.05225 (2017-11)
CheXNet: Radiologist-Level Pneumonia Detection on Chest X-Rays with Deep Learning
1804.07461 (2018-04)
GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
and 10 more not named here
entropy 1 other paper
1 dated earlier
2103.04505 (2021-03)
Split Computing and Early Exiting for Deep Learning Applications: Survey and Research Challenges

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

In this paper, we systematically evaluate the robustness of multi-exit language models against adversarial slowdown. To audit their robustness, we design a slowdown attack that generates natural adversarial text bypassing early-exit points. We use the resulting WAFFLE attack as a vehicle to conduct a comprehensive evaluation of three multi-exit mechanisms with the GLUE benchmark against adversarial slowdown. We then show our attack significantly reduces the computational savings provided by the three methods in both white-box and black-box settings. The more complex a mechanism is, the more vulnerable it is to adversarial slowdown. We also perform a linguistic analysis of the perturbed text inputs, identifying common perturbation patterns that our attack generates, and comparing them with standard adversarial text attacks. Moreover, we show that adversarial training is ineffective in defeating our slowdown attack, but input sanitization with a conversational model, e.g., ChatGPT, can remove perturbations effectively. This result suggests that future work is needed for developing efficient yet robust multi-exit models.

For agents

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

get_harvested_code_for_paper("2310.19152")
get_code_for_paper("2310.19152")
have("2310.19152")

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

curl https://syntology.ai/api/ran/2310.19152.json

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

[![Syntology run record](https://syntology.ai/api/ran/2310.19152.svg)](https://syntology.ai/paper/2310.19152)

Connect an agent — have() is free.