SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2603.13893 · 2026

UVLM: A Universal Vision-Language Model Loader for Reproducible Multimodal Benchmarking

Joan Perez, Giovanni Fusco

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 1 has no contract verdict; 2 failures of our instrument; 1 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/2603.13893.json · badge: /api/ran/2603.13893.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
perezjoan/UVLM community
Papers with Code archive
3 ran · 1 no recorded run
FunctionStatusWhere it lives
_resolve_vision_budget Ran
ran; our contract draft was wrong, not the code
perezjoan/UVLM/uvlm/loader.py
code served (permissive licence) · get_code("1338748ab9180ba7")
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, ndim 1, 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
import numpy as np

# Test case 1: None vision_budget
vision_budget_1 = None
backend_1 = 'qwen'

# Test case 2: 'native' vision_budget
vision_budget_2 = 'native'
backend_2 = 'llava'

# Test case 3: preset string 'low' for qwen
vision_budget_3 = 'low'
backend_3 = 'qwen'

# Test case 4: preset string 'medium' for internvl
vision_budget_4 = 'medium'
backend_4 = 'internvl'

# Test case 5: preset string 'high' for gemma4
vision_budget_5 = 'high'
backend_5 = 'gemma4'

# Test case 6: dict vision_budget for qwen3
vision_budget_6 = {'min_pixels': 1024, 'max_pixels': 2048}
backend_6 = 'qwen3'

# Test case 7: dict vision_budget for internvl
vision_budget_7 = {'crop_to_patches': False, 'min_patches': 2, 'max_patches': 4}
backend_7 = 'internvl'

# Test case 8: dict vision_budget for gemma4 with valid max_soft_tokens
vision_budget_8 = {'max_soft_tokens': 280}
backend_8 = 'gemma4'

# Test case 9: dict vision_budget for llava
vision_budget_9 = {'image_grid_pinpoints': [[336, 336], [672, 672]]}
backend_9 = 'llava'

# Test case 10: preset string 'low' for llava
vision_budget_10 = 'low'
backend_10 = 'llava'

ARGS = {
    'vision_budget': vision_budget_1,
    'backend': backend_1,
}
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 tuple, ndim 1, dtype family float — a tuple 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
get_hf_token Ran
ran; our contract draft was wrong, not the code
perezjoan/UVLM/uvlm/loader.py
code served (permissive licence) · get_code("0c8e5e8633262939")
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) 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 os

# The function get_hf_token takes one optional parameter: token
# It can be None or a string representing a HuggingFace token

# Since the function:
# 1. Checks if token is not None and returns it
# 2. Tries to get token from Colab secrets (if in Colab)
# 3. Tries to get token from HF_TOKEN environment variable
# 4. Tries to get token from huggingface_hub HfFolder
# 5. Returns None if not found
#
# We need to provide a valid argument for the 'token' parameter

# Test case 1: Explicit token provided
ARGS = {
    'token': 'hf_test_token_12345'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
google, huggingface_hub — 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
is_colab Ran
ran on a synthesized input
perezjoan/UVLM/uvlm/loader.py
code served (permissive licence) · get_code("f47214ec22ec1f25")
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) and returned bool, 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 os
import torch
import torch.nn as nn

torch.manual_seed(0)

# The function is_colab() takes no parameters
# It only checks for google.colab import and /content directory
# No arguments are needed

ARGS = {}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
google, 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 — no output shape is stated: observe_signature.py could not observe it (no_inputs)
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_model Not yet run
unverified (harvested, no recorded run)
perezjoan/UVLM/uvlm/loader.py
code served (permissive licence) · get_code("0ac9db5392bf70d0")
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

Vision-Language Models (VLMs) have emerged as powerful tools for image understanding tasks, yet their practical deployment remains hindered by significant architectural heterogeneity across model families. This paper introduces UVLM (Universal Vision-Language Model Loader), a Google Colab-based framework that provides a unified interface for loading, configuring, and benchmarking multiple VLM architectures on custom image analysis tasks. UVLM currently supports two major model families -LLaVA-NeXT and Qwen2.5-VLwhich differ fundamentally in their vision encoding, tokenization, and decoding strategies. The framework abstracts these differences behind a single inference function, enabling researchers to compare models using identical prompts and evaluation protocols. Key features include a multi-task prompt builder with support for four response types (numeric, category, boolean, text), a consensus validation mechanism based on majority voting across repeated inferences, a flexible token budget (up to 1,500 tokens) enabling users to design custom reasoning strategies through prompt engineering, and a built-in chain-of-thought reference mode for benchmarking. UVLM is designed for reproducibility, accessibility, and extensibility and as such is freely deployable on Google Colab using consumer-grade GPU resources. The paper also presents the first benchmarking of different VLMs on tasks of increasing reasoning complexity using a corpus of 120 street-view images.

For agents

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

get_harvested_code_for_paper("2603.13893")
get_code_for_paper("2603.13893")
have("2603.13893")

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

curl https://syntology.ai/api/ran/2603.13893.json

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

[![Syntology run record](https://syntology.ai/api/ran/2603.13893.svg)](https://syntology.ai/paper/2603.13893)

Connect an agent — have() is free.