SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2404.00909 · CVPR · 2024

Learning by Correction: Efficient Tuning Task for Zero-Shot Generative Vision-Language Reasoning

Yu Wu, Xuming He, Rongjie Li

arXiv · PDF · Open in the Atlas

Code that ran

8 of 10 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (1 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 2 violated it and 1 has no contract verdict; 4 failures of our instrument; 2 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/2404.00909.json · badge: /api/ran/2404.00909.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
shtuplus/iccc_cvpr2024 marked official in Papers with Code
Papers with Code archive
8 ran · 2 no recorded run
FunctionStatusWhere it lives
TreeNode Ran
ran on a synthesized input · constructed an object; no result was computed
shtuplus/iccc_cvpr2024/iccc_data_generation.py
code served (permissive licence) · get_code("fea778fe4e98da35")
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 an instance of its own class, TreeNode: the run constructed the object and did not call it, so no result was computed. 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 torch
import numpy as np
from unittest.mock import Mock

# Create a mock token object that TreeNode expects
mock_token = Mock()
mock_token.text = "example"
mock_token.dep_ = "nsubj"
mock_token.pos_ = "NOUN"

ARGS = {
    "idx": 0,
    "curr": mock_token,
    "ent_idx": None,
    "rel_idx": None
}
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 TreeNode — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a TreeNode 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
__flatten_conjunction Ran
ran; our contract draft was wrong, not the code
shtuplus/iccc_cvpr2024/iccc_data_generation.py
code served (permissive licence) · get_code("6710f1dc8352f98d")
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 generator. 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 torch
import torch.nn as nn

class MockNode:
    """Mock spacy-like node with children and dep_ attribute"""
    def __init__(self, dep_='ROOT', children=None):
        self.dep_ = dep_
        self.children = children if children is not None else []

# Create a simple tree structure
# Root node with some children, some marked as 'conj' (conjunctions)
child1 = MockNode(dep_='nsubj', children=[])
child2 = MockNode(dep_='conj', children=[])
child3 = MockNode(dep_='conj', children=[])
child4 = MockNode(dep_='verb', children=[])

root_node = MockNode(dep_='ROOT', children=[child1, child2, child3, child4])

ARGS = {
    'node': root_node
}
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 generator — a generator output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a generator 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
bfs Ran
ran; our contract draft was wrong, not the code
shtuplus/iccc_cvpr2024/iccc_data_generation.py
code served (permissive licence) · get_code("d7775e9e08c81ed2")
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 Node. 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.

class Node:
    def __init__(self, value, children_token=None):
        self.value = value
        self.children_token = children_token if children_token is not None else []

# Create a simple tree structure for BFS to traverse
node_leaf_1 = Node(value=5, children_token=[])
node_leaf_2 = Node(value=10, children_token=[])
node_leaf_3 = Node(value=15, children_token=[])

node_intermediate_1 = Node(value=3, children_token=[node_leaf_1, node_leaf_2])
node_intermediate_2 = Node(value=7, children_token=[node_leaf_3])

start_node = Node(value=1, children_token=[node_intermediate_1, node_intermediate_2])

# Define a match condition function that looks for a node with value == 10
def match_cond(node):
    return node.value == 10

ARGS = {
    "start_node": start_node,
    "match_cond": match_cond
}
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 Node — a Node output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Node 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
get_all_dep_ent Ran
ran; our contract draft was wrong, not the code
shtuplus/iccc_cvpr2024/iccc_data_generation.py
code served (permissive licence) · get_code("5d89fd1cce6bef72")
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 queue

class Token:
    def __init__(self, token_id, dep, children=None):
        self.id = token_id
        self.dep = dep
        self.children_token = children if children is not None else []

# Create a tree structure for testing
token_4 = Token(4, 'punct', [])
token_3 = Token(3, 'amod', [])
token_2 = Token(2, 'nsubj', [token_3, token_4])
token_1 = Token(1, 'root', [token_2])
start_node = token_1

ARGS = {
    'start_node': start_node,
    'degree': -1
}
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 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
get_all_dep_pred Ran
ran; our contract draft was wrong, not the code
shtuplus/iccc_cvpr2024/iccc_data_generation.py
code served (permissive licence) · get_code("39810fc192fe3ce1")
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 queue

class Node:
    def __init__(self, node_id, pos='NOUN', dep='nsubj', children_token=None):
        self.id = node_id
        self.pos = pos
        self.dep = dep
        self.children_token = children_token if children_token is not None else []

# Create a simple tree structure of nodes
child1 = Node(1, pos='VERB', dep='nmod')
child2 = Node(2, pos='ADP', dep='case')
child3 = Node(3, pos='NOUN', dep='nsubj')
child4 = Node(4, pos='PUNCT', dep='punct')

grandchild1 = Node(5, pos='NOUN', dep='obj')
grandchild2 = Node(6, pos='VERB', dep='acl')

child1.children_token = [grandchild1, grandchild2]
child2.children_token = []
child3.children_token = []
child4.children_token = []

start_node = Node(0, pos='VERB', dep='ROOT', children_token=[child1, child2, child3, child4])

ARGS = {
    'start_node': start_node,
    'degree': -1
}
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 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
get_node_id Ran
ran, honoured the contract we drafted
shtuplus/iccc_cvpr2024/iccc_data_generation.py
code served (permissive licence) · get_code("09cad294d3001f80")
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 int, ndim 0, dtype family int, 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

class Node:
    def __init__(self, node_id):
        self.id = node_id

node = Node(42)

ARGS = {
    "node": node
}
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 int, ndim 0, dtype family int, 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
is_obj Ran
ran, violated the contract we drafted
shtuplus/iccc_cvpr2024/iccc_data_generation.py
code served (permissive licence) · get_code("de90d155211d60bc")
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.

class Node:
    def __init__(self, dep, pos):
        self.dep = dep
        self.pos = pos

# Create test nodes with different dependency and POS tag combinations
node_with_obj = Node(dep='obj', pos='NOUN')

ARGS = {
    'node': node_with_obj
}
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 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
is_sub Ran
ran, violated the contract we drafted
shtuplus/iccc_cvpr2024/iccc_data_generation.py
code served (permissive licence) · get_code("5e0f3cfd75ecd33d")
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.

class Node:
    def __init__(self, dep):
        self.dep = dep

# Test case 1: node with 'sub' in dep string
node_with_sub = Node('sub')

# Test case 2: node without 'sub' in dep string
node_without_sub = Node('other')

ARGS = {
    'node': node_with_sub
}
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 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
ParsingTree Not yet run
unverified (harvested, no recorded run)
shtuplus/iccc_cvpr2024/iccc_data_generation.py
code served (permissive licence) · get_code("bbcfee4ec5ce7322")
How it ran

no invocation recorded.

get_entity Not yet run
unverified (harvested, no recorded run)
shtuplus/iccc_cvpr2024/iccc_data_generation.py
code served (permissive licence) · get_code("76e740a0c5bf1801")
How it ran

no invocation recorded.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 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 10 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

Generative vision-language models (VLMs) have shown impressive performance in zero-shot vision-language tasks like image captioning and visual question answering. However, improving their zero-shot reasoning typically requires second-stage instruction tuning, which relies heavily on human-labeled or large language model-generated annotation, incurring high labeling costs. To tackle this challenge, we introduce Image-Conditioned Caption Correction (ICCC), a novel pre-training task designed to enhance VLMs' zero-shot performance without the need for labeled task-aware data. The ICCC task compels VLMs to rectify mismatches between visual and language concepts, thereby enhancing instruction following and text generation conditioned on visual inputs. Leveraging language structure and a lightweight dependency parser, we construct data samples of ICCC task from image-text datasets with low labeling and computation costs. Experimental results on BLIP-2 and InstructBLIP demonstrate significant improvements in zero-shot image-text generation-based VL tasks through ICCC instruction tuning.

For agents

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

get_harvested_code_for_paper("2404.00909")
get_code_for_paper("2404.00909")
have("2404.00909")

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

curl https://syntology.ai/api/ran/2404.00909.json

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

[![Syntology run record](https://syntology.ai/api/ran/2404.00909.svg)](https://syntology.ai/paper/2404.00909)

Connect an agent — have() is free.