SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2602.11157 · NeurIPS · 2025

Response-Based Knowledge Distillation for Multilingual Jailbreak Prevention Unwittingly Compromises Safety

Kai Zhang, Kevin Zhu, Max Zhang, Derek Liu, Joshua Franco, Haihao Liu

arXiv · PDF · Open in the Atlas

Code that ran

4 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; 3 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 1 failure of our instrument; 0 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/2602.11157.json · badge: /api/ran/2602.11157.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
maxh119Z/RB-KD-Multilingual-Safety-Trade-offs canonical (classifier label)
classifier basis: census
4 ran · 0 no recorded run
FunctionStatusWhere it lives
compute_equation_acc Ran
ran, honoured the contract we drafted
maxh119Z/RB-KD-Multilingual-Safety-Trade-offs/metrics.py
code served (permissive licence) · get_code("8ecff480ccfa4b55")
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 float64, ndim 0, 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 numpy as np

def eval_equation(equation):
    try:
        answer = eval(equation)
    except:
        answer = np.nan
    return answer

def compute_equation_acc(preds, labels):
    preds = [eval_equation(pred) for pred in preds]
    labels = [eval_equation(label) for label in labels]
    return np.mean(np.array(preds) == np.array(labels))

# Setup code to construct valid arguments
preds = [
    "2 + 2",
    "3 * 4",
    "10 / 2",
    "5 - 1"
]

labels = [
    "4",
    "12",
    "5.0",
    "4"
]

ARGS = {
    "preds": preds,
    "labels": labels
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
compute_text_acc Ran
ran, honoured the contract we drafted
maxh119Z/RB-KD-Multilingual-Safety-Trade-offs/metrics.py
code served (permissive licence) · get_code("32f0f000c61d1f6b")
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 float64, ndim 0, 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 numpy as np

# Construct valid arguments for compute_text_acc
# The function takes preds and labels, converts them to numpy arrays,
# and computes element-wise equality mean

preds = [0, 1, 1, 0, 1]
labels = [0, 1, 0, 0, 1]

ARGS = {
    "preds": preds,
    "labels": labels,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
eval_equation Ran
ran, honoured the contract we drafted
maxh119Z/RB-KD-Multilingual-Safety-Trade-offs/metrics.py
code served (permissive licence) · get_code("343ca53bbef3f75c")
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 numpy as np

# The function eval_equation takes a single parameter: equation
# It evaluates the equation string and returns the result or np.nan if evaluation fails

# We'll create a simple mathematical equation string that can be safely evaluated
equation_str = "2 + 3 * 4"

ARGS = {
    "equation": equation_str
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — 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
get_config_dir Ran
ran; our contract draft was wrong, not the code
maxh119Z/RB-KD-Multilingual-Safety-Trade-offs/train_utils.py
code served (permissive licence) · get_code("c061c5b9bacdd262")
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

class Args:
    def __init__(self):
        self.dataset = "cifar10"
        self.from_pretrained = "meta-llama/Llama-2-7b-hf"
        self.lora_r = 8
        self.lora_alpha = 16

ARGS = {"args": Args()}
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

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, 3 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 0 later) and 1 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
compute_equation_acc 2 other papers
2 dated earlier
2305.02301 (2023-05)
Distilling Step-by-Step! Outperforming Larger Language Models with Less Training Data and Smaller Model Sizes
2025.emnlp-main.1418 (2025 (year only))
Synth-SBDH: A Synthetic Dataset of Social and Behavioral Determinants of Health for Clinical Text
compute_text_acc 2 other papers
2 dated earlier
2305.02301 (2023-05)
Distilling Step-by-Step! Outperforming Larger Language Models with Less Training Data and Smaller Model Sizes
2025.emnlp-main.1418 (2025 (year only))
Synth-SBDH: A Synthetic Dataset of Social and Behavioral Determinants of Health for Clinical Text
eval_equation 2 other papers
2 dated earlier
2305.02301 (2023-05)
Distilling Step-by-Step! Outperforming Larger Language Models with Less Training Data and Smaller Model Sizes
2025.emnlp-main.1418 (2025 (year only))
Synth-SBDH: A Synthetic Dataset of Social and Behavioral Determinants of Health for Clinical Text

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

Large language models (LLMs) are increasingly deployed worldwide, yet their safety alignment remains predominantly English-centric. This allows for vulnerabilities in non-English contexts, especially with low-resource languages. We introduce a novel application of knowledge distillation (KD) in the context of multilingual jailbreak prevention, examining its efficacy. We distill the refusal behaviors of a proprietary teacher model (OpenAI o1-mini) with Low-Rank Adaptation (LoRA) into three open-source student models: Meta-Llama-3-8B-Instruct, Gemma-2-2B-IT, and Qwen3-8B, using ∼28,000 multilingual jailbreak prompts from XSAFETY via black-box response-based, parameter-efficient fine-tuning (PEFT). Evaluation on the MULTIJAIL benchmark reveals a counterintuitive behavior: standard fine-tuning on the teacher's "safe" refusal data inadvertently increases Jailbreak Success Rate (JSR) for all student models, up to 16.6 percentage points. Our experiments reveal a divergent generalization to unseen languages during distillation, with varying outcomes depending on the base model. By removing a primary source of safety degradation, nuanced 'boundary' refusals, we mitigate or even reverse safety declines in student models, although reductions in reasoning performance (GSM8K) persist. Overall, our exploratory study highlights the challenges and potential of KD as a technique for multilingual safety alignment, offering a foundation for future research in this direction. Code is available at https://github.com/maxh119Z/RB-KD-Multilingual-Safety-Trade-offs.git

For agents

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

get_harvested_code_for_paper("2602.11157")
get_code_for_paper("2602.11157")
have("2602.11157")

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

curl https://syntology.ai/api/ran/2602.11157.json

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

[![Syntology run record](https://syntology.ai/api/ran/2602.11157.svg)](https://syntology.ai/paper/2602.11157)

Connect an agent — have() is free.