SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2402.07270 · ICLR · 2024

Open-ended VQA benchmarking of Vision-Language models by exploiting Classification datasets and their semantic hierarchy

Simon Ging, María Bravo

arXiv · PDF · Open in the Atlas

Code that ran

2 of 2 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; 2 failures of our instrument; 0 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/2402.07270.json · badge: /api/ran/2402.07270.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
lmb-freiburg/ovqa canonical (classifier label)
classifier basis: census
2 ran · 0 no recorded run
FunctionStatusWhere it lives
apply_layer_decay Ran
ran; our contract draft was wrong, not the code
lmb-freiburg/ovqa/ovqa/optim_layer_decay.py
code served (permissive licence) · get_code("3cd4a50601da36c4")
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 torch.nn as nn

torch.manual_seed(0)

class SimpleVisionTransformer(nn.Module):
    def __init__(self):
        super().__init__()
        # Base components
        self.visual_encoder_cls_token = nn.Parameter(torch.randn(1, 4))
        self.visual_encoder_pos_embed = nn.Parameter(torch.randn(1, 8, 4))
        self.visual_encoder_patch_embed_weight = nn.Parameter(torch.randn(4, 3))
        
        # Blocks with numbered layers
        self.visual_encoder_blocks_0_norm = nn.LayerNorm(4)
        self.visual_encoder_blocks_0_attn = nn.Linear(4, 4)
        self.visual_encoder_blocks_1_norm = nn.LayerNorm(4)
        self.visual_encoder_blocks_1_attn = nn.Linear(4, 4)
        self.visual_encoder_blocks_2_norm = nn.LayerNorm(4)
        self.visual_encoder_blocks_2_attn = nn.Linear(4, 4)
        
        # Other parameters
        self.head = nn.Linear(4, 2)
        
        # A frozen parameter
        self.frozen_param = nn.Parameter(torch.randn(3, 4))
        self.frozen_param.requires_grad = False

model = SimpleVisionTransformer()

ARGS = {
    'model': model,
    'lr_layer_decay': 0.9,
    'lr_layer_decay_regex_base': [
        r'visual_encoder_cls_token',
        r'visual_encoder_pos_embed',
        r'visual_encoder_patch_embed_weight'
    ],
    'lr_layer_decay_regex_layers': [
        r'visual_encoder_blocks_([0-9]+)_.*'
    ]
}
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 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
create_prompt Ran
ran; our contract draft was wrong, not the code
lmb-freiburg/ovqa/ovqa/create_prompt.py
code served (permissive licence) · get_code("e1cf91cb1d73f4b7")
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 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 logging

# Setup for create_prompt function
# The function expects:
# 1. samples: Dict[str, Any] - may contain 'text_input' key
# 2. prompt: str - template string with optional {} placeholder
# 3. batch_size: int - size of batch

# Test case 1: samples with text_input as list of strings
samples_case1 = {
    'text_input': ['What is AI?', 'How does ML work?'],
    'other_field': 'some_value'
}

prompt_case1 = 'Answer the following question: {}'
batch_size_case1 = 2

# Test case 2: samples with text_input as list of lists of strings
samples_case2 = {
    'text_input': [['Q1a', 'Q1b'], ['Q2a', 'Q2b']],
    'metadata': 'test'
}

prompt_case2 = 'Q: {} A:'
batch_size_case2 = 2

# Test case 3: samples without text_input, just use prompt template
samples_case3 = {
    'other_data': [1, 2, 3],
    'features': 'some_features'
}

prompt_case3 = 'This is a static prompt'
batch_size_case3 = 3

# Choose one case for ARGS (case 1 is simplest and most direct)
ARGS = {
    'samples': samples_case1,
    'prompt': prompt_case1,
    'batch_size': batch_size_case1
}
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 1, 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

Also present in other papers' code

Of 2 samples lifted from this paper's linked repositories, 2 carry a function identity and 0 do not. Of the 2 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 2 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

The evaluation of text-generative vision-language models is a challenging yet crucial endeavor. By addressing the limitations of existing Visual Question Answering (VQA) benchmarks and proposing innovative evaluation methodologies, our research seeks to advance our understanding of these models' capabilities. We propose a novel VQA benchmark based on well-known visual classification datasets which allows a granular evaluation of text-generative vision-language models and their comparison with discriminative vision-language models. To improve the assessment of coarse answers on fine-grained classification tasks, we suggest using the semantic hierarchy of the label space to ask automatically generated follow-up questions about the ground-truth category. Finally, we compare traditional NLP and LLM-based metrics for the problem of evaluating model predictions given ground-truth answers. We perform a human evaluation study upon which we base our decision on the final metric. We apply our benchmark to a suite of vision-language models and show a detailed comparison of their abilities on object, action, and attribute classification. Our contributions aim to lay the foundation for more precise and meaningful assessments, facilitating targeted progress in the exciting field of vision-language modeling.

For agents

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

get_harvested_code_for_paper("2402.07270")
get_code_for_paper("2402.07270")
have("2402.07270")

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

curl https://syntology.ai/api/ran/2402.07270.json

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

[![Syntology run record](https://syntology.ai/api/ran/2402.07270.svg)](https://syntology.ai/paper/2402.07270)

Connect an agent — have() is free.