SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2408.14512 · 2024

LLMs as Zero-shot Graph Learners: Alignment of GNN Representations with LLM Token Embeddings

arXiv · PDF · Open in the Atlas

Code that ran

9 of 9 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 8 have no contract verdict; 1 failure 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/2408.14512.json · badge: /api/ran/2408.14512.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
w-rudder/tea-glm canonical (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
MindSpore-scientific-2/code-11 community (classifier label)
classifier basis not recorded on the link
5 ran · 0 no recorded run
MindSpore-scientific-2/code-2 community (classifier label)
classifier basis not recorded on the link
3 ran · 0 no recorded run
FunctionStatusWhere it lives
create_dataset_info Ran
ran on a synthesized input
MindSpore-scientific-2/code-2/STCrowd/STCrowd_convert.py
code served (permissive licence) · get_code("665b410fa2e90913")
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 json
from pathlib import Path
import tempfile
import os

# Create a temporary directory structure for testing
temp_dir = tempfile.mkdtemp()
data_root = Path(temp_dir)
anno_dir = data_root / 'anno'
anno_dir.mkdir(parents=True, exist_ok=True)

# Create sample JSON files for groups
group_data_1 = {
    'total_number': 2,
    'frames': [
        {
            'frame_name': '/frame_0.pcd',
            'images': [{'image_name': '/frame_0.png'}],
            'items': [
                {
                    'id': 'obj_1',
                    'position': {'x': 1.0, 'y': 2.0, 'z': 1.5},
                    'boundingbox': {'x': 0.5, 'y': 0.6, 'z': 1.5},
                    'occlusion': 0,
                    'rotation': 0.1
                },
                {
                    'id': 'obj_2',
                    'position': {'x': 3.0, 'y': 4.0, 'z': 0.8},
                    'boundingbox': {'x': 0.7, 'y': 0.8, 'z': 0.8},
                    'occlusion': 1,
                    'rotation': 0.2
                }
            ]
        },
        {
            'frame_name': '/frame_1.pcd',
            'images': [{'image_name': '/frame_1.png'}],
            'items': [
                {
                    'id': 'obj_1',
                    'position': {'x': 1.2, 'y': 2.1, 'z': 1.5},
                    'boundingbox': {'x': 0.5, 'y': 0.6, 'z': 1.5},
                    'occlusion': 0,
                    'rotation': 0.11
                }
            ]
        }
    ]
}

group_data_2 = {
    'total_number': 1,
    'frames': [
        {
            'frame_name': '/frame_2.pcd',
            'images': [{'image_name': '/frame_2.png'}],
            'items': [
                {
                    'id': 'obj_3',
                    'position': {'x': 5.0, 'y': 6.0, 'z': 2.0},
                    'boundingbox': {'x': 0.8, 'y': 0.9, 'z': 2.0},
                    'occlusion': 0,
                    'rotation': 0.3
                }
            ]
        }
    ]
}

# Write group files
with open(anno_dir / 'group_0.json', 'w') as f:
    json.dump(group_data_1, f)

with open(anno_dir / 'group_1.json', 'w') as f:
    json.dump(group_data_2, f)

# Create dummy data files referenced in JSON
(data_root / 'frame_0.pcd').touch()
(data_root / 'frame_1.pcd').touch()
(data_root / 'frame_2.pcd').touch()
(data_root / 'frame_0.png').touch()
(data_root / 'frame_1.png').touch()
(data_root / 'frame_2.png').touch()

# Define arguments for create_dataset_info
ARGS = {
    'data_root': data_root,
    'split_list': ['group_0', 'group_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
decode Ran
ran on a synthesized input
MindSpore-scientific-2/code-11/AECRNet/data.py
code served (permissive licence) · get_code("72814d7de33e5aef")
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 Image. 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 numpy as np
from PIL import Image

# Create a small valid image array (Height=4, Width=3, Channels=3 for RGB)
# PIL.Image.fromarray expects a numpy array with shape (H, W) for grayscale
# or (H, W, 3) for RGB or (H, W, 4) for RGBA
img_array = np.random.randint(0, 256, size=(4, 3, 3), dtype=np.uint8)

ARGS = {
    "img": img_array
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 and dtype family int 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
generate_sample Ran
ran on a synthesized input
MindSpore-scientific-2/code-11/QuantExplainNLP_AlgoAnalysis/data_preprocessing.py
code served (permissive licence) · get_code("16298c972234921e")
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 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 numpy as np

class StandInTokenizer:
    """Minimal tokenizer stand-in"""
    def __init__(self, vocab_size):
        self.vocab_size = vocab_size
    
    def encode(self, text):
        return [np.random.randint(0, self.vocab_size) for _ in text.split()]
    
    def decode(self, tokens):
        return ' '.join(str(t) for t in tokens)

# Set seed for reproducibility
np.random.seed(42)

seq_length = 50

medical_terms = [
    'hypertension', 'diabetes', 'pneumonia', 'sepsis', 
    'trauma', 'stroke', 'myocardial', 'infection', 'fever', 'pain'
]

whole_word_tokens = [
    'patient', 'hospital', 'doctor', 'nurse', 'treatment',
    'diagnosis', 'symptom', 'disease', 'condition', 'care'
]

templates = [
    "The {term1} with {severity} {term2} required {treatment}.",
    "Patient presented with {term1} and {term2}, {outcome} after {treatment}.",
    "Clinical evaluation of {term1} in {term3} showed {severity} {term4}.",
    "The {term1} {term2} was managed with {treatment}, resulting in {outcome}.",
    "Following {term1}, the {term3} developed {severity} {term4} and received {treatment}.",
]

vocab_size = 100

tokenizer = StandInTokenizer(vocab_size=vocab_size)

ARGS = {
    "seq_length": seq_length,
    "medical_terms": medical_terms,
    "whole_word_tokens": whole_word_tokens,
    "templates": templates,
    "vocab_size": vocab_size,
    "tokenizer": tokenizer,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — 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
get_feature_names Ran
ran on a synthesized input
MindSpore-scientific-2/code-11/QuantExplainNLP_AlgoAnalysis/explainability.py
code served (permissive licence) · get_code("5919a74fc358ee01")
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 ndarray, ndim 1, dtype family float, shape (13,) 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.

from sklearn.feature_extraction.text import TfidfVectorizer

# Small sample texts for the vectorizer to fit on
texts = [
    "machine learning is great",
    "natural language processing",
    "text classification tasks",
    "feature extraction methods"
]

# tfidf_data parameter - this appears to be unused in the function body,
# but we'll provide a placeholder anyway
tfidf_data = None

# max_features parameter
max_features = 1407

ARGS = {
    "tfidf_data": tfidf_data,
    "texts": texts,
    "max_features": max_features
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
sklearn — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family float, shape (13,) — shape (13,) 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
get_title Ran
ran; our contract draft was wrong, not the code
w-rudder/tea-glm/GNNPreTrain/train_valid_paper.py
code served (permissive licence) · get_code("2ede4c92b35f0844")
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 dict. 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 tempfile
import os

# Create a temporary file with sample data
temp_file = tempfile.NamedTemporaryFile(mode='w', delete=False, suffix='.txt', encoding='utf-8')
temp_path = temp_file.name

# Write header and sample data
temp_file.write('paper_id\ttitle\tcontent\n')
temp_file.write('1\tTitle One\tContent One\n')
temp_file.write('2\tTitle Two\tContent Two\n')
temp_file.write('3\tTitle Three\tContent Three\n')
temp_file.write('5\tTitle Five\tContent Five\n')

# Add many lines to reach near line 179719 (only add enough to test the skip logic)
for i in range(4, 100):
    temp_file.write(f'{i}\tTitle {i}\tContent {i}\n')

temp_file.close()

# Create paperid2node mapping
paperid2node = {
    1: 'node_1',
    2: 'node_2',
    3: 'node_3',
    5: 'node_5',
}

# Add some more mappings for the extended lines
for i in range(4, 100):
    paperid2node[i] = f'node_{i}'

ARGS = {
    'path': temp_path,
    'paperid2node': paperid2node,
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
load_group Ran
ran on a synthesized input
MindSpore-scientific-2/code-2/STCrowd/STCrowd_convert.py
code served (permissive licence) · get_code("6553960c96b0e666")
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 json
import tempfile
from pathlib import Path

# Create a temporary directory to use as data_root
temp_dir = tempfile.TemporaryDirectory()
data_root = Path(temp_dir.name)

# Create necessary subdirectories
(data_root / "images").mkdir(exist_ok=True)
(data_root / "point_clouds").mkdir(exist_ok=True)

# Create a temporary group file with valid JSON structure
group_file_content = {
    "total_number": 2,
    "frames": [
        {
            "frame_name": "/point_clouds/frame_0.pcd",
            "images": [
                {
                    "image_name": "/images/image_0.jpg"
                }
            ],
            "items": [
                {
                    "id": "obj_1",
                    "position": {"x": 1.0, "y": 2.0, "z": 3.0},
                    "boundingbox": {"x": 0.5, "y": 0.6, "z": 1.5},
                    "occlusion": 0,
                    "rotation": 0.5
                },
                {
                    "id": "obj_2",
                    "position": {"x": 4.0, "y": 5.0, "z": 0.8},
                    "boundingbox": {"x": 0.4, "y": 0.5, "z": 1.2},
                    "occlusion": 1,
                    "rotation": 1.2
                }
            ]
        },
        {
            "frame_name": "/point_clouds/frame_1.pcd",
            "images": [
                {
                    "image_name": "/images/image_1.jpg"
                }
            ],
            "items": [
                {
                    "id": "obj_1",
                    "position": {"x": 1.1, "y": 2.1, "z": 3.1},
                    "boundingbox": {"x": 0.5, "y": 0.6, "z": 1.5},
                    "occlusion": 0,
                    "rotation": 0.5
                },
                {
                    "id": "obj_3",
                    "position": {"x": 5.0, "y": 6.0, "z": 1.3},
                    "boundingbox": {"x": 0.3, "y": 0.4, "z": 1.1},
                    "occlusion": 2,
                    "rotation": 0.8
                }
            ]
        }
    ]
}

# Write the group file
group_file = data_root / "group.json"
with open(group_file, 'w') as f:
    json.dump(group_file_content, f)

# Create dummy files so file paths exist
(data_root / "images" / "image_0.jpg").touch()
(data_root / "images" / "image_1.jpg").touch()
(data_root / "point_clouds" / "frame_0.pcd").touch()
(data_root / "point_clouds" / "frame_1.pcd").touch()

ARGS = {
    "data_root": data_root,
    "group_file": group_file
}
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
maybe_num_nodes Ran
ran on a synthesized input
MindSpore-scientific-2/code-11/RGNN/model_simple.py
code served (permissive licence) · get_code("ef93260b5137f73b")
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

torch.manual_seed(0)

# The function calls index.max().item() + 1
# So index needs to be a tensor with a max() method
index = torch.tensor([0, 1, 2, 3, 4], dtype=torch.long)

# num_nodes is optional; we'll test both cases
num_nodes = None

ARGS = {
    "index": index,
    "num_nodes": num_nodes
}
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
process_single_scene Ran
ran on a synthesized input
MindSpore-scientific-2/code-2/STCrowd/STCrowd_convert.py
code served (permissive licence) · get_code("da6a321dfe042d21")
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 dict. 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.

from pathlib import Path

# Construct load_dict with proper structure
load_dict = {
    'frames': [
        {
            'frame_name': '/frame_0.pcd',
            'images': [{'image_name': '/image_0.jpg'}],
            'items': [
                {
                    'id': 'obj_1',
                    'position': {'x': 1.0, 'y': 2.0, 'z': 3.0},
                    'boundingbox': {'x': 0.5, 'y': 0.6, 'z': 1.5},
                    'occlusion': 0,
                    'rotation': 0.5
                },
                {
                    'id': 'obj_2',
                    'position': {'x': 5.0, 'y': 6.0, 'z': 0.8},
                    'boundingbox': {'x': 0.4, 'y': 0.5, 'z': 1.2},
                    'occlusion': 1,
                    'rotation': 1.2
                }
            ]
        },
        {
            'frame_name': '/frame_1.pcd',
            'images': [{'image_name': '/image_1.jpg'}],
            'items': [
                {
                    'id': 'obj_1',
                    'position': {'x': 1.5, 'y': 2.5, 'z': 3.2},
                    'boundingbox': {'x': 0.5, 'y': 0.6, 'z': 1.5},
                    'occlusion': 0,
                    'rotation': 0.6
                },
                {
                    'id': 'obj_3',
                    'position': {'x': 7.0, 'y': 8.0, 'z': 2.0},
                    'boundingbox': {'x': 0.7, 'y': 0.8, 'z': 1.4},
                    'occlusion': 0,
                    'rotation': 0.3
                }
            ]
        }
    ]
}

data_root = Path('/tmp/mock_data')

ARGS = {
    'load_dict': load_dict,
    'idx': 1,
    'data_root': data_root
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
set_random_seed Ran
ran on a synthesized input
MindSpore-scientific-2/code-11/AECRNet/data.py
code served (permissive licence) · get_code("14c181fabe865cbf")
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 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 random
import torch

torch.manual_seed(0)

ARGS = {
    "img_name": "test_image.jpg",
    "seed": 42
}
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 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

Also present in other papers' code

Of 9 samples lifted from this paper's linked repositories, 9 carry a function identity and 0 do not. Of the 9 distinct functions with an identity, 5 are also present in other papers whose code we hold (5 with at least one paper dated earlier, 0 the same month, 5 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.

FunctionAlso present inEarliest-datedOthers named
maybe_num_nodes 9 other papers
5 dated earlier, 4 later
2104.01528 (2021-04)
SGCN:Sparse Graph Convolution Network for Pedestrian Trajectory Prediction
2210.12624 (2022-10)
Mitigating Gradient Bias in Multi-objective Learning: A Provably Convergent Stochastic Approach
2307.16372 (2023-07)
LP-MusicCaps: LLM-Based Pseudo Music Captioning
and 6 more not named here
generate_sample 6 other papers
3 dated earlier, 3 later
2307.16372 (2023-07)
LP-MusicCaps: LLM-Based Pseudo Music Captioning
2312.16476 (2023-12)
SVGDreamer: Text Guided SVG Generation with Diffusion Model
2403.16372 (2024-03)
SignSGD with Federated Voting
and 3 more not named here
get_feature_names 6 other papers
3 dated earlier, 3 later
2307.16372 (2023-07)
LP-MusicCaps: LLM-Based Pseudo Music Captioning
2312.16476 (2023-12)
SVGDreamer: Text Guided SVG Generation with Diffusion Model
2403.16372 (2024-03)
SignSGD with Federated Voting
and 3 more not named here
decode 2 other papers
1 dated earlier, 1 later
2311.18815 (2023-11)
IMMA: Immunizing text-to-image Models against Malicious Adaptation
2503.09573 (2025-03)
BLOCK DIFFUSION: INTERPOLATING BETWEEN AU-TOREGRESSIVE AND DIFFUSION LANGUAGE MODELS
set_random_seed 2 other papers
1 dated earlier, 1 later
2311.18815 (2023-11)
IMMA: Immunizing text-to-image Models against Malicious Adaptation
2503.09573 (2025-03)
BLOCK DIFFUSION: INTERPOLATING BETWEEN AU-TOREGRESSIVE AND DIFFUSION LANGUAGE MODELS

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

Zero-shot graph machine learning, especially with graph neural networks (GNNs), has garnered significant interest due to the challenge of scarce labeled data. While methods like self-supervised learning and graph prompt learning have been extensively explored, they often rely on fine-tuning with task-specific labels, limiting their effectiveness in zero-shot scenarios. Inspired by the zero-shot capabilities of instruction-fine-tuned large language models (LLMs), we introduce a novel framework named Token Embedding-Aligned Graph Language Model (TEA-GLM) that leverages LLMs as cross-dataset and cross-task zero-shot learners for graph machine learning. Concretely, we pretrain a GNN, aligning its representations with token embeddings of an LLM. We then train a linear projector that transforms the GNN's representations into a fixed number of graph token embeddings without tuning the LLM. A unified instruction is designed for various graph tasks at different levels, such as node classification (node-level) and link prediction (edge-level). These design choices collectively enhance our method's effectiveness in zero-shot learning, setting it apart from existing methods. Experiments show that our graph token embeddings help the LLM predictor achieve state-of-the-art performance on unseen datasets and tasks compared to other methods using LLMs as predictors.

For agents

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

get_harvested_code_for_paper("2408.14512")
get_code_for_paper("2408.14512")
have("2408.14512")

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

curl https://syntology.ai/api/ran/2408.14512.json

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

[![Syntology run record](https://syntology.ai/api/ran/2408.14512.svg)](https://syntology.ai/paper/2408.14512)

Connect an agent — have() is free.