SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2408.04449 · 2024

EARBench: Towards Evaluating Physical Risk Awareness for Task Planning of Foundation Model-based Embodied AI Agents

arXiv · PDF · Open in the Atlas

Code that ran

6 of 11 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; 6 failures of our instrument; 5 no recorded run), measured 2026-11-21; 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/2408.04449.json · badge: /api/ran/2408.04449.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
zihao-ai/eairiskbench canonical (classifier label)
classifier basis: census
6 ran · 5 no recorded run
FunctionStatusWhere it lives
encode_image Ran
ran; our contract draft was wrong, not the code
zihao-ai/eairiskbench/image_judger.py
pointer only (licence: MIT); the stored copy get_code serves, from jongwoopark7978/LVNet, is pointer only (licence: NONE) · get_code("f41cb1a19b154297")
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 jongwoopark7978/LVNet, 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.

llm_generate Ran
ran; our contract draft was wrong, not the code
zihao-ai/eairiskbench/utils/llm_util.py
code served (permissive licence) · get_code("a903b8a603cdfbf1")
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

torch.manual_seed(0)

class MockLLMModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 8)
    
    def generate(self, user_prompt, system_prompt, response_format=None):
        # Mock implementation that returns a string response
        if response_format is not None:
            return f"Response in format: {response_format}"
        return "This is a generated response."

model = MockLLMModel()
user_prompt = "What is the capital of France?"
system_prompt = "You are a helpful assistant."
response_format = None

ARGS = {
    'model': model,
    'user_prompt': user_prompt,
    'system_prompt': system_prompt,
    'response_format': response_format
}
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
llm_generate_with_image Ran
ran; our contract draft was wrong, not the code
zihao-ai/eairiskbench/utils/llm_util.py
code served (permissive licence) · get_code("ba772f40b4fca257")
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
from PIL import Image
import io

torch.manual_seed(0)

class MockImageGenerativeModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 8)
    
    def generate_with_image(self, image_path, user_prompt, system_prompt=None, response_format=None):
        return f"Generated response for prompt: {user_prompt}"

model = MockImageGenerativeModel()

# Create a small valid image file in memory
img = Image.new('RGB', (4, 4), color='red')
img_byte_arr = io.BytesIO()
img.save(img_byte_arr, format='PNG')
image_data = img_byte_arr.getvalue()

user_prompt = "What is in this image?"
system_prompt = "You are a helpful assistant."
response_format = None

ARGS = {
    "model": model,
    "image_path": "dummy_image.png",
    "user_prompt": user_prompt,
    "system_prompt": system_prompt,
    "response_format": response_format,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, 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
llm_response_to_json Ran
ran; our contract draft was wrong, not the code
zihao-ai/eairiskbench/utils/llm_util.py
code served (permissive licence) · get_code("54c4e7d565d15e51")
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
import re
import ast

# Test case 1: Direct JSON string
response_1 = '{"key": "value", "number": 42}'

# Test case 2: JSON in markdown code block
response_2 = '''```json
{"name": "test", "status": "active"}
'''

# Test case 3: JSON in generic code block
response_3 = '''
{"result": "success", "count": 10}
'''

# Test case 4: JSON with braces extracted from text
response_4 = 'Here is the result: {"extracted": true, "data": "found"} at the end'

# Test case 5: Comma-separated key:value pairs
response_5 = 'name: Alice, age: 30, city: NYC'

# Test case 6: Response with newlines that will be stripped
response_6 = '''{"multi": "line",
"json": "object"}'''

ARGS = {
    "response": response_1
}
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
scene2observation Ran
ran; our contract draft was wrong, not the code
zihao-ai/eairiskbench/text_observation_generation.py
code served (permissive licence) · get_code("c71bbca6b41b4f93")
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 MockLanguageModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(10, 10)
    
    def generate(self, user_prompt, system_prompt, max_tokens, response_format):
        # Mock implementation that returns a simple string observation
        return "You are in a room. There is a table with objects on it. The room is well-lit."


torch.manual_seed(0)

scene = "A kitchen with a counter and dining area"
objects = ["microwave_1", "counter_1", "table_1", "bowl_1", "sandwich_1"]
object_positions = {
    "microwave_1": "on counter_1",
    "counter_1": "in kitchen",
    "table_1": "next to counter_1",
    "bowl_1": "on table_1",
    "sandwich_1": "on table_1"
}
object_attributes = {
    "microwave_1": ["closed door", "stainless steel"],
    "counter_1": ["marble", "kitchen"],
    "table_1": ["wooden", "dining"],
    "bowl_1": ["metal", "stainless steel"],
    "sandwich_1": ["fresh"]
}
model = MockLanguageModel()

ARGS = {
    "scene": scene,
    "objects": objects,
    "object_positions": object_positions,
    "object_attributes": object_attributes,
    "model": model
}
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
text2img_prompt_generator Ran
ran; our contract draft was wrong, not the code
zihao-ai/eairiskbench/text2image_prompt_generation.py
code served (permissive licence) · get_code("3fa1f780e19467e2")
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

torch.manual_seed(0)

class MockLanguageModel(nn.Module):
    """Mock language model that has a generate method."""
    
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(10, 10)
    
    def generate(self, user_prompt, system_prompt, max_tokens, response_format):
        """Mock generate method that returns a simple prompt string."""
        return "A realistic living room with a wooden table on the left side and a red sofa on the right side. A painting hangs on the back wall. Modern lighting from ceiling fixtures. Clean, well-organized space with neutral colors."

scene = "A modern living room"
objects = ["wooden table", "red sofa", "painting", "ceiling light"]
object_positions = {
    "wooden table": "left side of the room",
    "red sofa": "right side of the room",
    "painting": "back wall center",
    "ceiling light": "center of ceiling"
}
object_attributes = {
    "wooden table": "dark wood, rectangular, modern style",
    "red sofa": "bright red fabric, 3-seater, contemporary",
    "painting": "abstract art, gold frame",
    "ceiling light": "warm white LED, recessed"
}
model = MockLanguageModel()

ARGS = {
    "scene": scene,
    "objects": objects,
    "object_positions": object_positions,
    "object_attributes": object_attributes,
    "model": model
}
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
encode_image_resize Not yet run
unverified (harvested, no recorded run)
zihao-ai/eairiskbench/llm_models/gpt.py
code served (permissive licence) · get_code("de1417761254cb61")
How it ran

no invocation recorded.

evaluate_image_with_text Not yet run
unverified (harvested, no recorded run)
zihao-ai/eairiskbench/image_judger.py
code served (permissive licence) · get_code("993f45a996de6344")
How it ran

no invocation recorded.

request_task_status Not yet run
unverified (harvested, no recorded run)
zihao-ai/eairiskbench/scene_image_generation.py
code served (permissive licence) · get_code("1763eac67135cc1e")
How it ran

no invocation recorded.

save_img_from_url Not yet run
unverified (harvested, no recorded run)
zihao-ai/eairiskbench/scene_image_generation.py
code served (permissive licence) · get_code("24e1a09cc2fe1c1b")
How it ran

no invocation recorded.

submit_midjounery_task Not yet run
unverified (harvested, no recorded run)
zihao-ai/eairiskbench/scene_image_generation.py
code served (permissive licence) · get_code("e3f440fea8d5e620")
How it ran

no invocation recorded.

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 1 the same month, 1 later) and 10 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
encode_image 76 other papers
26 dated earlier, 2 the same month, 47 later, 1 same year, order not determinable
2309.05950 (2023-09)
Language Models as Black-Box Optimizers for Vision-Language Models
2311.15732 (2023-11)
GPT4Vis: What Can GPT-4 Do for Zero-shot Visual Recognition?
2312.14867 (2023-12)
VIEScore: Towards Explainable Metrics for Conditional Image Synthesis Evaluation
and 73 more not named here

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 2026-11-21.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Embodied artificial intelligence (EAI) integrates advanced AI models into physical entities for real-world interaction. The emergence of foundation models as the "brain" of EAI agents for high-level task planning has shown promising results. However, the deployment of these agents in physical environments presents significant safety challenges. For instance, a housekeeping robot lacking sufficient risk awareness might place a metal container in a microwave, potentially causing a fire. To address these critical safety concerns, comprehensive pre-deployment risk assessments are imperative. This study introduces EARBench, a novel framework for automated physical risk assessment in EAI scenarios. EAIRiskBench employs a multi-agent cooperative system that leverages various foundation models to generate safety guidelines, create risk-prone scenarios, make task planning, and evaluate safety systematically. Utilizing this framework, we construct EARDataset, comprising diverse test cases across various domains, encompassing both textual and visual scenarios. Our comprehensive evaluation of state-of-the-art foundation models reveals alarming results: all models exhibit high task risk rates (TRR), with an average of 95.75% across all evaluated models. To address these challenges, we further propose two prompting-based risk mitigation strategies. While these strategies demonstrate some efficacy in reducing TRR, the improvements are limited, still indicating substantial safety concerns. This study provides the first large-scale assessment of physical risk awareness in EAI agents. Our findings underscore the critical need for enhanced safety measures in EAI systems and provide valuable insights for future research directions in developing safer embodied artificial intelligence system. Data and code are available at https://github.com/zihao-ai/EARBench.

For agents

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

get_harvested_code_for_paper("2408.04449")
get_code_for_paper("2408.04449")
have("2408.04449")

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

curl https://syntology.ai/api/ran/2408.04449.json

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

[![Syntology run record](https://syntology.ai/api/ran/2408.04449.svg)](https://syntology.ai/paper/2408.04449)

Connect an agent — have() is free.