SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2402.09063 · 2024

Soft Prompt Threats: Attacking Safety Alignment and Unlearning in Open-Source LLMs through the Embedding Space

arXiv · PDF · Open in the Atlas

Code that ran

1 of 6 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; 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/2402.09063.json · badge: /api/ran/2402.09063.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
schwinnl/llm_embedding_attack canonical (classifier label)
classifier basis: census
1 ran · 5 no recorded run
FunctionStatusWhere it lives
create_one_hot Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
schwinnl/llm_embedding_attack/unlearning_utils.py
code served (permissive licence) · get_code("0339945f3e408b02")
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 3, dtype family float, shape (2, 3, 5) 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 transformers import LlamaForCausalLM
from unittest.mock import Mock

torch.manual_seed(0)

# Create a minimal mock LlamaForCausalLM model
model = Mock(spec=LlamaForCausalLM)
model.device = torch.device('cpu')

# Create embedding matrix with vocab_size=5, embedding_dim=4
embed_weights = torch.randn(5, 4, dtype=torch.float32)
model.model = Mock()
model.model.embed_tokens = Mock()
model.model.embed_tokens.weight = embed_weights

# Create token tensor: batch_size=2, sequence_length=3
tokens = torch.tensor([[1, 2, 0], [3, 1, 4]], dtype=torch.long)

# Make isinstance work for LlamaForCausalLM
model.__class__ = LlamaForCausalLM

ARGS = {
    'model': model,
    'tokens': tokens
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, transformers — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (2, 3, 5) — shape (2, 3, 5) 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
create_one_hot_and_embeddings Not yet run
unverified (harvested, no recorded run)
schwinnl/llm_embedding_attack/unlearning_utils.py
code served (permissive licence) · get_code("0e1764f3a0787969")
How it ran

no invocation recorded.

generate Not yet run
unverified (harvested, no recorded run)
schwinnl/llm_embedding_attack/embedding_attack_toxic.py
code served (permissive licence) · get_code("8b6b7c037edb3ff3")
How it ran

no invocation recorded.

get_embedding_matrix Not yet run
unverified (harvested, no recorded run)
schwinnl/llm_embedding_attack/embedding_attack_toxic.py
code served (permissive licence) · get_code("688aaeba506e3329")
How it ran

no invocation recorded.

init_attack_embeddings Not yet run
unverified (harvested, no recorded run)
schwinnl/llm_embedding_attack/unlearning_utils.py
code served (permissive licence) · get_code("3461002fd537de34")
How it ran

no invocation recorded.

load_model_and_tokenizer Not yet run
unverified (harvested, no recorded run)
schwinnl/llm_embedding_attack/embedding_attack_toxic.py
code served (permissive licence) · get_code("067aa4bb86ba8911")
How it ran

no invocation recorded.

Also present in other papers' code

Of 6 samples lifted from this paper's linked repositories, 6 carry a function identity and 0 do not. Of the 6 distinct functions with an identity, 2 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 1 the same month, 2 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.

FunctionAlso present inEarliest-datedOthers named
get_embedding_matrix 6 other papers
1 the same month, 5 later
2402.16459 (2024-02)
Defending LLMs against Jailbreaking Attacks via Backtranslation
2403.04957 (2024-03)
Automatic and Universal Prompt Injection Attacks against Large Language Models
2404.13968 (2024-04)
Protecting Your LLMs with Information Bottleneck
and 3 more not named here
load_model_and_tokenizer 1 other paper
1 later
2410.04234 (2024-10)
Functional Homotopy: Smoothing Discrete Optimization via Continuous Parameters for LLM Jailbreak Attacks

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

Current research in adversarial robustness of LLMs focuses on discrete input manipulations in the natural language space, which can be directly transferred to closed-source models. However, this approach neglects the steady progression of open-source models. As open-source models advance in capability, ensuring their safety also becomes increasingly imperative. Yet, attacks tailored to open-source LLMs that exploit full model access remain largely unexplored. We address this research gap and propose the embedding space attack, which directly attacks the continuous embedding representation of input tokens. We find that embedding space attacks circumvent model alignments and trigger harmful behaviors more efficiently than discrete attacks or model fine-tuning. Furthermore, we present a novel threat model in the context of unlearning and show that embedding space attacks can extract supposedly deleted information from unlearned LLMs across multiple datasets and models. Our findings highlight embedding space attacks as an important threat model in open-source LLMs. Trigger Warning: the appendix contains LLM-generated text with violence and harassment.

For agents

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

get_harvested_code_for_paper("2402.09063")
get_code_for_paper("2402.09063")
have("2402.09063")

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

curl https://syntology.ai/api/ran/2402.09063.json

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

[![Syntology run record](https://syntology.ai/api/ran/2402.09063.svg)](https://syntology.ai/paper/2402.09063)

Connect an agent — have() is free.