SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2407.02273 · NeurIPS · 2024

Language Model Alignment in Multilingual Trolley Problems

Bernhard Schölkopf, Yejin Choi, Mrinmaya Sachan, Max Kleiman-Weiner, Zhijing Jin, Sydney Levine, Fernando Gonzalez, Rada Mihalcea, Jiarui Liu, Giorgio Piatti, Francesco Ortu, András Strausz

arXiv · PDF · Open in the Atlas

Code that ran

3 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; 3 failures of our instrument; 1 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/2407.02273.json · badge: /api/ran/2407.02273.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
causalNLP/moralmachine canonical (classifier label)
classifier basis: census
3 ran · 1 no recorded run
FunctionStatusWhere it lives
get_model_release_date Ran
ran; our contract draft was wrong, not the code
causalNLP/moralmachine/multi_tp/models_ids.py
code served (permissive licence) · get_code("51b748076654186d")
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 numpy as np

# The function get_model_release_date takes a single parameter 'model'
# which is a string key into the llm_models dictionary
# It returns llm_models[model]['release_date']

# Valid model keys from the llm_models dict include:
# MISTRAL_7B, LLAMA_3_1_8B, LLAMA_3_1_70B, LLAMA_3_8B, LLAMA_3_70B,
# LLAMA_2_70B, LLAMA_2_13B, LLAMA_2_7B, QWEN_2_72B, QWEN_2_7B,
# PHI_3_MEDIUM, PHI_3_5_MINI, PHI_3_5_MOE, GEMMA_2_9B, GEMMA_2_2B,
# GEMMA_2_27B, GPT_4_OMNI_MINI, GPT_4, GPT_3

MISTRAL_7B = 'mistralai/Mistral-7B-Instruct-v0.2'
LLAMA_3_1_8B = 'meta-llama/Meta-Llama-3.1-8B-Instruct'
LLAMA_3_1_70B = 'neuralmagic/Meta-Llama-3.1-70B-Instruct-FP8'
LLAMA_3_8B = 'meta-llama/Meta-Llama-3-8B-Instruct'
LLAMA_3_70B = 'neuralmagic/Meta-Llama-3-70B-Instruct-FP8'
LLAMA_2_70B = 'meta-llama/Llama-2-70b-chat-hf'
LLAMA_2_13B = 'meta-llama/Llama-2-13b-chat-hf'
LLAMA_2_7B = 'meta-llama/Llama-2-7b-chat-hf'
QWEN_2_7B = 'Qwen/Qwen2-7B-Instruct'
QWEN_2_72B = 'Qwen/Qwen2-72B-Instruct-GPTQ-Int8'
PHI_3_MEDIUM = 'microsoft/Phi-3-medium-4k-instruct'
PHI_3_5_MINI = 'microsoft/Phi-3.5-mini-instruct'
PHI_3_5_MOE = 'microsoft/Phi-3.5-MoE-instruct'
GEMMA_2_9B = 'google/gemma-2-9b-it'
GEMMA_2_2B = 'google/gemma-2b-it'
GEMMA_2_27B = 'google/gemma-2-27b-it'
GPT_4_OMNI_MINI = 'gpt-4o-mini-2024-07-18'
GPT_4 = 'gpt-4-0613'
GPT_3 = 'text-davinci-003'

ARGS = {
    'model': MISTRAL_7B
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, 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
get_size Ran
ran; our contract draft was wrong, not the code
causalNLP/moralmachine/multi_tp/models_ids.py
code served (permissive licence) · get_code("bf351864d0daa92c")
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 numpy as np

# The function get_size(model) takes a model string as parameter
# and returns llm_models[model]['size']
# So we need to provide a valid model key from the llm_models dictionary

ARGS = {
    'model': 'mistralai/Mistral-7B-Instruct-v0.2'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, 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
parse_dict_from_json Ran
ran; our contract draft was wrong, not the code
causalNLP/moralmachine/code/run_llama.py
code served (permissive licence) · get_code("04ea8f490ac507f8")
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 dict. 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 json

def parse_dict_from_json(json_str):
    try:
        return json.loads(json_str)
    except json.JSONDecodeError:
        return {}

# Valid JSON string that will parse successfully
valid_json_str = '{"key": "value", "number": 42, "nested": {"inner": true}}'

# Invalid JSON string that will trigger JSONDecodeError
invalid_json_str = '{invalid json}'

# Setup ARGS with both test cases
ARGS = {
    "json_str": valid_json_str
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — 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
get_family Not yet run
unverified (harvested, no recorded run)
causalNLP/moralmachine/multi_tp/models_ids.py
code served (permissive licence) · get_code("bf7c01cef3555b12")
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

We evaluate the moral alignment of LLMs with human preferences in multilingual trolley problems. Building on the Moral Machine experiment, which captures over 40 million human judgments across 200+ countries, we develop a crosslingual corpus of moral dilemma vignettes in over 100 languages called MUL-TITP. This dataset enables the assessment of LLMs' decision-making processes in diverse linguistic contexts. Our analysis explores the alignment of 19 different LLMs with human judgments, capturing preferences across six moral dimensions: species, gender, fitness, status, age, and the number of lives involved. By correlating these preferences with the demographic distribution of language speakers and examining the consistency of LLM responses to various prompt paraphrasings, our findings provide insights into cross-lingual and ethical biases of LLMs and their intersection. We discover significant variance in alignment across languages, challenging the assumption of uniform moral reasoning in AI systems and highlighting the importance of incorporating diverse perspectives in AI ethics. The results underscore the need for further research on the integration of multilingual dimensions in responsible AI research to ensure fair and equitable AI interactions worldwide. 1

For agents

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

get_harvested_code_for_paper("2407.02273")
get_code_for_paper("2407.02273")
have("2407.02273")

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

curl https://syntology.ai/api/ran/2407.02273.json

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

[![Syntology run record](https://syntology.ai/api/ran/2407.02273.svg)](https://syntology.ai/paper/2407.02273)

Connect an agent — have() is free.