SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2503.02199 · CVPR · 2025

Words or Vision: Do Vision-Language Models Have Blind Faith in Text?

Bryan Hooi, Ailin Deng, Tri Cao, Zhirui Chen

arXiv · PDF · Open in the Atlas

Code that ran

3 of 3 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, 1 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/2503.02199.json · badge: /api/ran/2503.02199.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
d-ailin/blind-faith-in-text canonical (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
FunctionStatusWhere it lives
check_answer_match Ran
ran, violated the contract we drafted
d-ailin/blind-faith-in-text/raw_code/hf_evaluator.py
code served (permissive licence) · get_code("20931aa6c8bc7cbc")
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, 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
import numpy as np

# Analyze the function check_answer_match
# Parameters:
# - pred: a prediction (string or convertible to string)
# - gt_answers: a list/collection of ground truth answers (strings or convertible)

# The function calls:
# - str(pred) - converts pred to string
# - str(gt) for each gt in gt_answers - converts each to string
# - normalize_answer() on both
# - string comparison operations (==, in)

# Build test arguments
pred_value = "two"
gt_answers_value = ["2", "two", "three"]

ARGS = {
    "pred": pred_value,
    "gt_answers": gt_answers_value
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, 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
normalize_answer Ran
ran; our contract draft was wrong, not the code
d-ailin/blind-faith-in-text/raw_code/hf_evaluator.py
code served (permissive licence) · get_code("0bcff406512311b4")
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 re

torch_imported = False
try:
    import torch
    torch_imported = True
except ImportError:
    pass

# Test cases for normalize_answer function
test_cases = [
    "The answer is twenty three",
    "One hundred and fifty",
    "Hello, world!",
    "UPPERCASE TEXT",
    "   spaces   everywhere   ",
    42,
    "zero one two three four five",
    "What is the answer? It's forty-two!",
]

ARGS = {
    "answer": "The answer is twenty three"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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
parse_string_list Ran
ran; our contract draft was wrong, not the code
d-ailin/blind-faith-in-text/raw_code/hf_evaluator.py
code served (permissive licence) · get_code("171a0e48f4bc4eb0")
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 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 ast

# Test case 1: Already a list
val1 = ['answer1', 'answer2']

# Test case 2: String representation of a list
val2 = "['item1', 'item2', 'item3']"

# Test case 3: String representation of a list with numbers
val3 = "[1, 2, 3]"

# Test case 4: Plain string (not a list format)
val4 = "single_answer"

# Test case 5: Non-string, non-list value
val5 = 42

# Test case 6: Empty list string
val6 = "[]"

# Test case 7: String with brackets but invalid syntax
val7 = "[unclosed"

ARGS = {
    "val": val1
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 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 3 samples lifted from this paper's linked repositories, 3 carry a function identity and 0 do not. Of the 3 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 3 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) excel in integrating visual and textual information for vision-centric tasks, but their handling of inconsistencies between modalities is underexplored. We investigate VLMs' modality preferences when faced with visual data and varied textual inputs in visioncentered settings. By introducing textual variations to four vision-centric tasks and evaluating ten Vision-Language Models (VLMs), we discover a "blind faith in text" phenomenon: VLMs disproportionately trust textual data over visual data when inconsistencies arise, leading to significant performance drops under corrupted text and raising safety concerns. We analyze factors influencing this text bias, including instruction prompts, language model size, text relevance, token order, and the interplay between visual and textual certainty. While certain factors, such as scaling up the language model size, slightly mitigate text bias, others like token order can exacerbate it due to positional biases inherited from language models. To address this issue, we explore supervised fine-tuning with text augmentation and demonstrate its effectiveness in reducing text bias. Additionally, we provide a theoretical analysis suggesting that the blind faith in text phenomenon may stem from an imbalance of pure text and multi-modal data during training. Our findings highlight the need for balanced training and careful consideration of modality interactions in VLMs to enhance their robustness and reliability in handling multi-modal data inconsistencies.

For agents

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

get_harvested_code_for_paper("2503.02199")
get_code_for_paper("2503.02199")
have("2503.02199")

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

curl https://syntology.ai/api/ran/2503.02199.json

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

[![Syntology run record](https://syntology.ai/api/ran/2503.02199.svg)](https://syntology.ai/paper/2503.02199)

Connect an agent — have() is free.