SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2412.07012 · 2024

ProVision: Programmatically Scaling Vision-centric Instruction Data for Multimodal Language Models

arXiv · PDF · Open in the Atlas

Code that ran

7 of 15 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; 1 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 6 failures of our instrument; 8 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/2412.07012.json · badge: /api/ran/2412.07012.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
jieyuz2/provision canonical (classifier label)
classifier basis: census
7 ran · 8 no recorded run
FunctionStatusWhere it lives
boxOverlap Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
jieyuz2/provision/provision/dataset.py
code served (permissive licence) · get_code("1b9b3ba31d9f08f9")
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 bool, 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)

# box1 and box2 are expected to be sequences with 4 elements: [x1, y1, x2, y2]
# where (x1, y1) is top-left and (x2, y2) is bottom-right
box1 = [10, 20, 30, 40]
box2 = [25, 35, 45, 55]

ARGS = {
    "box1": box1,
    "box2": box2,
    "vertical": False,
    "horizontal": False,
}
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 bool, 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
load_image Ran
ran, honoured the contract we drafted
jieyuz2/provision/provision/annotation/attribute_llava.py
code served (permissive licence) · get_code("5f90221712cb63f4")
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 Image. 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 tempfile
from PIL import Image
import os

# Create a temporary image file for testing
temp_dir = tempfile.mkdtemp()
temp_image_path = os.path.join(temp_dir, "test_image.jpg")

# Create a simple test image (RGB, small size)
test_image = Image.new('RGB', (4, 3), color=(255, 0, 0))
test_image.save(temp_image_path)

# Define arguments dict
ARGS = {
    'image_file': temp_image_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, requests — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
normalize_relations Ran
ran; our contract draft was wrong, not the code
jieyuz2/provision/provision/dataset.py
code served (permissive licence) · get_code("3cfade61c1d29398")
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 list, ndim 2, 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 torch

torch.manual_seed(0)

EXCLUDE_RELATIONS = ['of']

def normalize_relations(relations):
    if relations is None:
        return None
    return [r for r in relations if r[1] not in EXCLUDE_RELATIONS]

# relations parameter should be a list of tuples where each tuple has at least 2 elements
# and the second element (r[1]) is checked against EXCLUDE_RELATIONS
relations_input = [
    ('subject', 'of', 'object'),
    ('agent', 'does', 'action'),
    ('thing', 'of', 'another'),
    ('entity', 'has', 'property')
]

ARGS = {
    'relations': relations_input
}
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 list, ndim 2, dtype family float — a list output has no array shape of its own; none is stated
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
rel_parsing Ran
ran; our contract draft was wrong, not the code
jieyuz2/provision/provision/annotation/relation_grounding.py
code served (permissive licence) · get_code("a1e26f63b595bca0")
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 list, ndim 2, 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 re
from typing import Union

# Test case 1: holistic detection mode with properly formatted string
holistic_input = """
Some header text
Relations:
region1 to region2, region3: interacts with, contains
region2 to region1, region4: above, next to
"""

# Test case 2: detailed detection mode with list of strings
detailed_input = [
    "interacts with region2, contains region3",
    "above region1, next to region4",
    "overlaps region5"
]

ARGS = {
    'rel_output': holistic_input,
    'detection_mode': 'holistic'
}
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 list, ndim 2, dtype family float — a list output has no array shape of its own; none is stated
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
save_depth_res Ran
ran; our contract draft was wrong, not the code
jieyuz2/provision/provision/annotation/depth_estimation.py
code served (permissive licence) · get_code("9008ca38b586fa68")
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 numpy as np
import tempfile
import os

# Create a temporary directory for saving
temp_dir = tempfile.mkdtemp()
save_path = os.path.join(temp_dir, "depth.npy")

# Create sample depth data - a 2D array with varying values
depth_data = np.array([
    [0.1, 0.5, 0.9, 1.2],
    [0.3, 0.7, 1.1, 1.5],
    [0.2, 0.6, 1.0, 1.3]
])

ARGS = {
    "image_size": (3, 4),
    "depth": depth_data,
    "save_path": save_path,
    "grayscale": False
}
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 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
save_mask_as_png Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
jieyuz2/provision/provision/annotation/utils.py
code served (permissive licence) · get_code("b67cc1d3454644b8")
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 numpy as np
import tempfile
import os

# Create a temporary directory for output
temp_dir = tempfile.mkdtemp()

# Create sample masks as numpy arrays (2D binary masks)
mask1 = np.zeros((4, 3), dtype=np.uint8)
mask1[0:2, 0:2] = 1

mask2 = np.zeros((4, 3), dtype=np.uint8)
mask2[2:4, 1:3] = 1

# Test with array-based masks
masks_array = [mask1, mask2]

# Test with dict-based masks (RLE format like COCO)
masks_dict = [
    {'segmentation': mask1.astype(np.uint8)},
    {'segmentation': mask2.astype(np.uint8)}
]

# Define output file path
output_file = os.path.join(temp_dir, "test_mask.png")

# Create ARGS for the function call using array-based masks
ARGS = {
    'masks': masks_array,
    'file_path': output_file,
    'grey_scale': False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — 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
save_sep_mask Ran
ran; our contract draft was wrong, not the code
jieyuz2/provision/provision/annotation/utils.py
code served (permissive licence) · get_code("2ac58005d7f320e4")
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 tuple. 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 pickle as pkl
import numpy as np
from scipy.sparse import csr_matrix
import tempfile
import os

# Create a temporary directory for saving
temp_dir = tempfile.mkdtemp()
save_path = os.path.join(temp_dir, "test_mask.pkl")

# Create small sample mask data
# Each mask is a dict with 'segmentation' and 'bbox'
mask1_segmentation = np.array([[1, 0, 1], [0, 1, 0], [1, 1, 0]], dtype=bool)
mask1_bbox = [10, 20, 30, 40]

mask2_segmentation = np.array([[0, 1], [1, 0]], dtype=bool)
mask2_bbox = [5, 15, 25, 35]

masks = [
    {
        'segmentation': mask1_segmentation,
        'bbox': mask1_bbox
    },
    {
        'segmentation': mask2_segmentation,
        'bbox': mask2_bbox
    }
]

ARGS = {
    'masks': masks,
    'save_path': save_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is 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
cate_grounding Not yet run
unverified (harvested, no recorded run)
jieyuz2/provision/provision/annotation/relation_grounding.py
code served (permissive licence) · get_code("7f28431ff5823dfe")
How it ran

no invocation recorded.

depth_estimation Not yet run
unverified (harvested, no recorded run)
jieyuz2/provision/provision/annotation/depth_estimation.py
code served (permissive licence) · get_code("7a269cd987117ade")
How it ran

no invocation recorded.

get_attributes Not yet run
unverified (harvested, no recorded run)
jieyuz2/provision/provision/annotation/attribute_llava.py
code served (permissive licence) · get_code("3010b213fcb650e6")
How it ran

no invocation recorded.

get_image Not yet run
unverified (harvested, no recorded run)
jieyuz2/provision/provision/annotation/utils.py
code served (permissive licence) · get_code("4354809003526c37")
How it ran

no invocation recorded.

normalize_attributes Not yet run
unverified (harvested, no recorded run)
jieyuz2/provision/provision/dataset.py
code served (permissive licence) · get_code("a72c327b29ab95e1")
How it ran

no invocation recorded.

object_detection Not yet run
unverified (harvested, no recorded run)
jieyuz2/provision/provision/annotation/object_detection.py
code served (permissive licence) · get_code("a13bc543209b74ab")
How it ran

no invocation recorded.

sam_mask_generation Not yet run
unverified (harvested, no recorded run)
jieyuz2/provision/provision/annotation/segmentation.py
code served (permissive licence) · get_code("f69b840efc02bd76")
How it ran

no invocation recorded.

sam_mask_prediction Not yet run
unverified (harvested, no recorded run)
jieyuz2/provision/provision/annotation/segmentation.py
code served (permissive licence) · get_code("8b3900eb71dae6ba")
How it ran

no invocation recorded.

Also present in other papers' code

Of 15 samples lifted from this paper's linked repositories, 15 carry a function identity and 0 do not. Of the 15 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 15 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

With the rise of multimodal applications, instruction data has become critical for training multimodal language models capable of understanding complex image-based queries. Existing practices rely on powerful but costly large language models (LLMs) or multimodal language models (MLMs) to produce instruction data. These are often prone to hallucinations, licensing issues and the generation process is often hard to scale and interpret. In this work, we present a programmatic approach that employs scene graphs as symbolic representations of images and human-written programs to systematically synthesize vision-centric instruction data. Our approach ensures the interpretability and controllability of the data generation process and scales efficiently while maintaining factual accuracy. By implementing a suite of 24 single-image, 14 multi-image instruction generators, and a scene graph generation pipeline, we build a scalable, cost-effective system: ProVision which produces diverse question-answer pairs concerning objects, attributes, relations, depth, etc., for any given image. Applied to Visual Genome and DataComp datasets, we generate over 10 million instruction data points, ProVision-10M, and leverage them in both pretraining and instruction tuning stages of MLMs. When adopted in the instruction tuning stage, our single-image instruction data yields up to a 7% improvement on the 2D split and 8% on the 3D split of CVBench, along with a 3% increase in performance on QBench2, RealWorldQA, and MMMU. Our multi-image instruction data leads to an 8% improvement on Mantis-Eval. Incorporation of our data in both pre-training and fine-tuning stages of xGen-MM-4B leads to an averaged improvement of 1.6% across 11 benchmarks.

For agents

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

get_harvested_code_for_paper("2412.07012")
get_code_for_paper("2412.07012")
have("2412.07012")

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

curl https://syntology.ai/api/ran/2412.07012.json

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

[![Syntology run record](https://syntology.ai/api/ran/2412.07012.svg)](https://syntology.ai/paper/2412.07012)

Connect an agent — have() is free.