SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2410.09421 · EMNLP · 2024

VLFeedback: A Large-Scale AI Feedback Dataset for Large Vision-Language Models Alignment

Lei Li, Yazheng Yang, Lingpeng Kong, Qi Liu, Peiyi Wang, Benyou Wang, Mukai Li, Liang, Zhihui Xie, Shunian Chen

arXiv · PDF · Open in the Atlas

Code that ran

5 of 7 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; 5 failures of our instrument; 2 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/2410.09421.json · badge: /api/ran/2410.09421.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
yuweihao/MM-Vet canonical (classifier label)
classifier basis: census
5 ran · 2 no recorded run
FunctionStatusWhere it lives
encode_image Ran
ran; our contract draft was wrong, not the code
yuweihao/MM-Vet/inference/utils.py
pointer only (licence: Apache-2.0); 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 (Apache-2.0), 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.

get_file_names Ran
ran; our contract draft was wrong, not the code
yuweihao/MM-Vet/mm-vet_evaluator.py
code served (permissive licence) · get_code("7d2672b31e9aa80e")
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 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 os

class Args:
    def __init__(self):
        self.gpt_model = "gpt-3.5"
        self.num_run = 5
        self.result_path = "/tmp/results"

ARGS = {
    "args": Args(),
    "model": "bert-base",
    "sub_set_name": "_validation"
}
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 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_file_names Ran
ran; our contract draft was wrong, not the code
yuweihao/MM-Vet/v2/mm-vet-v2_evaluator.py
code served (permissive licence) · get_code("a907f50fb10a0162")
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 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 os

class Args:
    def __init__(self):
        self.gpt_model = "gpt4"
        self.num_run = 5
        self.result_path = "/tmp/results"

ARGS = {
    "args": Args(),
    "model": "bert",
    "subset_name": "_test"
}
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 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
load_metadata Ran
ran; our contract draft was wrong, not the code
yuweihao/MM-Vet/mm-vet_evaluator.py
code served (permissive licence) · get_code("969f70cdb9e7b42d")
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 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 json
import os
import tempfile
import shutil

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

# Create mm-vet.json
mmvet_data = {
    "image_001": {
        "capability": ["ocr", "scene_understanding"]
    },
    "image_002": {
        "capability": ["object_detection", "scene_understanding"]
    },
    "image_003": {
        "capability": ["ocr", "object_detection", "scene_understanding"]
    },
    "image_004": {
        "capability": ["ocr"]
    }
}

mmvet_path = temp_dir
mmvet_file = os.path.join(mmvet_path, 'mm-vet.json')
with open(mmvet_file, 'w') as f:
    json.dump(mmvet_data, f)

# Create bard_set.json for subset testing
bard_set_data = ["image_001", "image_002", "image_003"]
bard_set_file = os.path.join(mmvet_path, 'bard_set.json')
with open(bard_set_file, 'w') as f:
    json.dump(bard_set_data, f)

# Create args object
class Args:
    def __init__(self, mmvet_path, use_sub_set):
        self.mmvet_path = mmvet_path
        self.use_sub_set = use_sub_set

ARGS = {
    'args': Args(mmvet_path=mmvet_path, use_sub_set=False)
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, pandas — 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
load_metadata Ran
ran; our contract draft was wrong, not the code
yuweihao/MM-Vet/v2/mm-vet-v2_evaluator.py
code served (permissive licence) · get_code("9385df01fd8cdfee")
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 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 json
import os
import tempfile
import pathlib

# Create temporary directory and files for testing
temp_dir = tempfile.mkdtemp()
mmvetv2_path = temp_dir

# Create mm-vet-v2.json with sample data
mmvet_data = {
    "image_001": {
        "capability": ["ocr", "scene_understanding"]
    },
    "image_002": {
        "capability": ["object_detection", "scene_understanding"]
    },
    "image_003": {
        "capability": ["ocr", "object_detection", "scene_understanding"]
    },
    "image_004": {
        "capability": ["ocr"]
    }
}

mmvet_json_path = os.path.join(mmvetv2_path, 'mm-vet-v2.json')
with open(mmvet_json_path, 'w') as f:
    json.dump(mmvet_data, f)

# Create subset.json with a subset of image IDs
subset_data = ["image_001", "image_002", "image_003"]
subset_path = os.path.join(temp_dir, 'subset.json')
with open(subset_path, 'w') as f:
    json.dump(subset_data, f)

# Create args object
class Args:
    def __init__(self, mmvetv2_path, subset=None):
        self.mmvetv2_path = mmvetv2_path
        self.subset = subset

# Test with subset
ARGS = {
    'args': Args(mmvetv2_path=mmvetv2_path, subset=subset_path)
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, pandas — 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
arg_parser Not yet run
unverified (harvested, no recorded run)
yuweihao/MM-Vet/mm-vet_evaluator.py
code served (permissive licence) · get_code("4edc6079065ecd0f")
How it ran

no invocation recorded.

arg_parser Not yet run
unverified (harvested, no recorded run)
yuweihao/MM-Vet/v2/mm-vet-v2_evaluator.py
code served (permissive licence) · get_code("436825beacc9df7d")
How it ran

no invocation recorded.

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 carry a function identity and 0 do not. Of the 7 distinct functions with an identity, 7 are also present in other papers whose code we hold (7 with at least one paper dated earlier, 1 the same month, 1 later) and 0 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
31 dated earlier, 12 the same month, 32 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
arg_parser 3 other papers
3 dated earlier
2308.02490 (2023-08)
MM-Vet: Evaluating Large Multimodal Models for Integrated Capabilities
2405.19716 (2024-05)
Enhancing Large Vision Language Models with Self-Training on Image Comprehension
2408.00765 (2024-08)
MM-Vet v2: A Challenging Benchmark to Evaluate Large Multimodal Models for Integrated Capabilities
get_file_names 3 other papers
3 dated earlier
2308.02490 (2023-08)
MM-Vet: Evaluating Large Multimodal Models for Integrated Capabilities
2405.19716 (2024-05)
Enhancing Large Vision Language Models with Self-Training on Image Comprehension
2408.00765 (2024-08)
MM-Vet v2: A Challenging Benchmark to Evaluate Large Multimodal Models for Integrated Capabilities
load_metadata 3 other papers
3 dated earlier
2308.02490 (2023-08)
MM-Vet: Evaluating Large Multimodal Models for Integrated Capabilities
2405.19716 (2024-05)
Enhancing Large Vision Language Models with Self-Training on Image Comprehension
2408.00765 (2024-08)
MM-Vet v2: A Challenging Benchmark to Evaluate Large Multimodal Models for Integrated Capabilities
arg_parser 1 other paper
1 dated earlier
2408.00765 (2024-08)
MM-Vet v2: A Challenging Benchmark to Evaluate Large Multimodal Models for Integrated Capabilities
get_file_names 1 other paper
1 dated earlier
2408.00765 (2024-08)
MM-Vet v2: A Challenging Benchmark to Evaluate Large Multimodal Models for Integrated Capabilities
load_metadata 1 other paper
1 dated earlier
2408.00765 (2024-08)
MM-Vet v2: A Challenging Benchmark to Evaluate Large Multimodal Models for Integrated Capabilities

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

As large vision-language models (LVLMs) evolve rapidly, the demand for high-quality and diverse data to align these models becomes increasingly crucial. However, the creation of such data with human supervision proves costly and time-intensive. In this paper, we investigate the efficacy of AI feedback to scale supervision for aligning LVLMs. We introduce VLFeedback, the first large-scale vision-language feedback dataset, comprising over 82K multi-modal instructions and comprehensive rationales generated by off-the-shelf models without human annotations. To evaluate the effectiveness of AI feedback for vision-language alignment, we train Silkie, an LVLM fine-tuned via direct preference optimization on VLFeedback. Silkie showcases exceptional performance regarding helpfulness, visual faithfulness, and safety metrics. It outperforms its base model by 6.9% and 9.5% in perception and cognition tasks, reduces hallucination issues on MMHal-Bench, and exhibits enhanced resilience against redteaming attacks. Furthermore, our analysis underscores the advantage of AI feedback, particularly in fostering preference diversity to deliver more comprehensive improvements. Our dataset, training code and models are available at https://vlf-silkie.github.io.

For agents

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

get_harvested_code_for_paper("2410.09421")
get_code_for_paper("2410.09421")
have("2410.09421")

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

curl https://syntology.ai/api/ran/2410.09421.json

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

[![Syntology run record](https://syntology.ai/api/ran/2410.09421.svg)](https://syntology.ai/paper/2410.09421)

Connect an agent — have() is free.