SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2503.08685 · ICCV · 2025

"Principal Components" Enable A New Language of Images

Xiaojuan Qi, Ismail Elezi, Xin Wen, Bingchen Zhao, Jiankang Deng

arXiv · PDF · Open in the Atlas

Code that ran

14 of 17 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; 4 honoured the contract we drafted, 0 violated it and 7 have no contract verdict; 3 failures of our instrument; 3 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/2503.08685.json · badge: /api/ran/2503.08685.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
visual-gen/semanticist canonical (classifier label)
classifier basis not recorded on the link
14 ran · 3 no recorded run
FunctionStatusWhere it lives
build_mlp Ran
ran; our contract draft was wrong, not the code
visual-gen/semanticist/semanticist/stage1/diffuse_slot.py
pointer only (licence: MIT); the stored copy get_code serves, from csslc/Self-Transcendence, is pointer only (licence: NONE) · get_code("b05a447cc3b9e991")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from csslc/Self-Transcendence, 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.

concat_all_gather Ran
ran on a synthesized input
visual-gen/semanticist/semanticist/engine/trainer_utils.py
code served (permissive licence) · get_code("d0aca56def6a1f55")
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 Tensor, ndim 2, dtype family float, shape (4, 4) 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 os

# Set random seed for reproducibility
torch.manual_seed(0)

# Mock torch.distributed functions since we're in a single-process environment
class MockDistributed:
    _world_size = 2
    
    @staticmethod
    def get_world_size():
        return MockDistributed._world_size
    
    @staticmethod
    def all_gather(tensor_list, tensor, async_op=False):
        # Simulate all_gather by filling the list with copies of the tensor
        for i in range(len(tensor_list)):
            tensor_list[i].copy_(tensor)

# Monkey patch torch.distributed for this test
torch.distributed.get_world_size = MockDistributed.get_world_size
torch.distributed.all_gather = MockDistributed.all_gather

# Create a small test tensor: shape (2, 4)
# This represents a batch of 2 samples with 4 features
tensor = torch.randn(2, 4)

ARGS = {
    "tensor": tensor
}
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 Tensor, ndim 2, dtype family float, shape (4, 4) — shape (4, 4) 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
decode_n_tokens Ran
ran on a synthesized input
visual-gen/semanticist/semanticist/stage2/generate.py
code served (permissive licence) · get_code("c10d644e68f17884")
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. 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

torch.manual_seed(0)

class MockModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 4)
    
    def forward(self, x, cond_idx=None, input_pos=None, cfg=1.0, temperature=1.0):
        # x shape: (batch, seq_len, 4) or (2*batch, seq_len, 4) when cfg > 1
        # Return a token tensor of shape (batch, 1) with token indices
        batch_size = x.shape[0]
        # Return token indices (e.g., values in range [0, 999])
        return torch.randint(0, 1000, (batch_size, 1))

model = MockModel()

# cur_token: shape (batch_size, 1) - the current token to start with
cur_token = torch.randint(0, 1000, (2, 1))

# input_pos: shape (1,) - current position in sequence
input_pos = torch.tensor([0])

# num_new_tokens: int - number of tokens to generate
num_new_tokens = 3

# cfg_scale: float - classifier-free guidance scale
cfg_scale = 1.5

# cfg_schedule: str - schedule type for cfg
cfg_schedule = 'constant'

# temperature: float - temperature for sampling
temperature = 1.0

ARGS = {
    'model': model,
    'cur_token': cur_token,
    'input_pos': input_pos,
    'num_new_tokens': num_new_tokens,
    'cfg_scale': cfg_scale,
    'cfg_schedule': cfg_schedule,
    'temperature': temperature,
}
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 — a list 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
decode_one_token Ran
ran on a synthesized input
visual-gen/semanticist/semanticist/stage2/generate.py
code served (permissive licence) · get_code("f6e7c115a93e25a8")
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 Tensor, ndim 1, dtype family int, shape (4,) 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 torch.nn as nn

torch.manual_seed(0)

class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.embedding = nn.Embedding(256, 4)
        self.linear = nn.Linear(4, 256)
    
    def forward(self, x, cond_idx=None, input_pos=None, cfg=1.0, temperature=1.0):
        # x shape: (batch, seq_len)
        batch_size = x.shape[0]
        seq_len = x.shape[1]
        
        # Embed tokens
        embedded = self.embedding(x)  # (batch, seq_len, 4)
        
        # Simple processing
        output = self.linear(embedded)  # (batch, seq_len, 256)
        
        # Return tokens for last position
        tokens = output[:, -1, :].argmax(dim=-1)  # (batch,)
        
        return tokens

model = SimpleModel()

# x tensor: batch of 2, sequence length of 3, token values in [0, 255]
x = torch.randint(0, 256, (2, 3), dtype=torch.long)

# input_pos tensor: shape (1,) indicating position
input_pos = torch.tensor([2], dtype=torch.long)

# cfg_scale: classifier-free guidance scale
cfg_scale = 1.5

# temperature: sampling temperature
temperature = 1.0

ARGS = {
    "model": model,
    "x": x,
    "input_pos": input_pos,
    "cfg_scale": cfg_scale,
    "temperature": temperature,
}
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 Tensor, ndim 1, dtype family int, shape (4,) — shape (4,) 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
drop_path Ran
ran on a synthesized input
visual-gen/semanticist/semanticist/stage1/vision_transformer.py
code served (permissive licence) · get_code("4134f673786a2abe")
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 Tensor, ndim 2, dtype family float, shape (2, 4) 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)

# Construct a small tensor (batch_size=2, feature_dim=4)
x = torch.randn(2, 4)

# Define arguments for drop_path function
ARGS = {
    "x": x,
    "drop_prob": 0.5,
    "training": True
}
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 Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) 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
find_multiple Ran
ran, honoured the contract we drafted
visual-gen/semanticist/semanticist/stage2/gpt.py
code served (permissive licence) · get_code("f6ff7671338c9c92")
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
import numpy as np

torch.manual_seed(0)

# Analyze the function: find_multiple(n: int, k: int)
# The function takes two integers and returns an integer
# It checks if n is divisible by k, and if not, rounds n up to the next multiple of k

ARGS = {
    "n": 10,
    "k": 3
}
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 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
get_1d_sincos_pos_embed Ran
ran on a synthesized input
visual-gen/semanticist/semanticist/stage1/pos_embed.py
code served (permissive licence) · get_code("e13eb3c1606755e6")
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 2, dtype family float, shape (4, 8) 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 numpy as np

# Construct arguments for get_1d_sincos_pos_embed
embed_dim = 8  # Must be even
grid_size = 4  # Number of positions to encode

ARGS = {
    "embed_dim": embed_dim,
    "grid_size": grid_size,
}
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 ndarray, ndim 2, dtype family float, shape (4, 8) — shape (4, 8) 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_2d_sincos_pos_embed Ran
ran, honoured the contract we drafted
visual-gen/semanticist/semanticist/stage1/diffusion_transfomer.py
code served (permissive licence) · get_code("c92c27c924b517e8")
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 ndarray, ndim 2, dtype family float, shape (18, 8) 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 numpy as np

# Define ARGS dict with all parameters for get_2d_sincos_pos_embed
ARGS = {
    "embed_dim": 8,
    "grid_size": 4,
    "cls_token": True,
    "extra_tokens": 2,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (18, 8) — shape (18, 8) 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_2d_sincos_pos_embed Ran
ran, honoured the contract we drafted
visual-gen/semanticist/semanticist/stage1/pos_embed.py
code served (permissive licence) · get_code("9417ae492629cf06")
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 2, dtype family float, shape (16, 8) 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 numpy as np

# The function get_2d_sincos_pos_embed takes:
# - embed_dim: int, must be even (embedding dimension)
# - grid_size: int, size of the grid (height and width)
# - cls_token: bool, whether to prepend a cls token

ARGS = {
    "embed_dim": 8,
    "grid_size": 4,
    "cls_token": False,
}
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 ndarray, ndim 2, dtype family float, shape (16, 8) — shape (16, 8) 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_obj_from_str Ran
ran; our contract draft was wrong, not the code
visual-gen/semanticist/semanticist/engine/trainer_utils.py
code served (permissive licence) · get_code("1be6845a690a8075")
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 type. 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 importlib
import torch
import torch.nn as nn

torch.manual_seed(0)

# The function get_obj_from_str takes a string path and optionally a reload flag
# It splits the string into module and class name, optionally reloads the module,
# then returns the class/object from that module

# We need to construct valid arguments:
# - string: a valid module path like "module.ClassName"
# - reload: a boolean

# Let's use a real example from torch
string_arg = "torch.nn.Linear"

# reload is a boolean parameter
reload_arg = False

ARGS = {
    "string": string_arg,
    "reload": reload_arg,
}
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 type — a type output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a type 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
interpolate_features Ran
ran on a synthesized input
visual-gen/semanticist/semanticist/stage1/diffuse_slot.py
code served (permissive licence) · get_code("1f5b50d685e20502")
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 Tensor, ndim 3, dtype family float, shape (2, 9, 4) 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 math

torch.manual_seed(0)

# The function takes a tensor x of shape (B, T1, D) where T1 must be a perfect square
# and target_len must also be a perfect square (since it's converted to H2, W2)
# Let's use T1 = 16 (4x4) and target_len = 9 (3x3)

B = 2  # batch size
T1 = 16  # sequence length (must be perfect square: 4*4)
D = 4  # feature dimension
target_len = 9  # target sequence length (must be perfect square: 3*3)

# Create a tensor of shape (B, T1, D)
x = torch.randn(B, T1, D)

ARGS = {
    "x": x,
    "target_len": target_len,
}
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 Tensor, ndim 3, dtype family float, shape (2, 9, 4) — shape (2, 9, 4) 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
modulate Ran
ran; our contract draft was wrong, not the code
visual-gen/semanticist/semanticist/stage1/diffusion_transfomer.py
code served (permissive licence) · get_code("03310bba324ae4fb")
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 Tensor, ndim 2, dtype family float, shape (2, 4) 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)

# Create small tensors for the modulate function
# x: input tensor with shape (batch_size, feature_dim)
x = torch.randn(2, 4)

# shift: tensor that will be unsqueezed to (batch_size, 1)
# so it should have shape (batch_size,)
shift = torch.randn(2)

# scale: tensor that will be unsqueezed to (batch_size, 1)
# so it should have shape (batch_size,)
scale = torch.randn(2)

ARGS = {
    "x": x,
    "shift": shift,
    "scale": scale,
}
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 Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) 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
modulate Ran
ran, honoured the contract we drafted
visual-gen/semanticist/semanticist/stage2/diffloss.py
code served (permissive licence) · get_code("62fcb3912a967a50")
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 Tensor, ndim 2, dtype family float, shape (2, 4) 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)

# Create small tensors for the modulate function
x = torch.randn(2, 4)  # batch size 2, dimension 4
shift = torch.randn(2, 4)  # same shape as x
scale = torch.randn(2, 4)  # same shape as x

ARGS = {
    "x": x,
    "shift": shift,
    "scale": scale,
}
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 Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) 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
prefill Ran
ran on a synthesized input
visual-gen/semanticist/semanticist/stage2/generate.py
code served (permissive licence) · get_code("d90c14200bec99a5")
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 Tensor, ndim 3, dtype family float, shape (2, 1, 8) 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 torch.nn as nn

torch.manual_seed(0)

class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 8)
    
    def forward(self, x, cond_idx, input_pos, cfg=1.0, temperature=1.0):
        # Dummy implementation that returns a tensor of shape (batch_size, seq_len, vocab_size)
        batch_size = cond_idx.shape[0]
        seq_len = cond_idx.shape[1] if cond_idx.dim() > 1 else 1
        # Return logits shape (batch_size, seq_len, vocab_size) but we'll reduce to (batch_size,)
        return torch.randn(batch_size, 8)

model = SimpleModel()

# cond_idx: conditioning indices tensor
cond_idx = torch.randint(0, 100, (2, 3))

# input_pos: input position tensor
input_pos = torch.arange(3).unsqueeze(0).expand(2, -1)

# cfg_scale: classifier-free guidance scale
cfg_scale = 1.5

# temperature: sampling temperature
temperature = 0.8

ARGS = {
    "model": model,
    "cond_idx": cond_idx,
    "input_pos": input_pos,
    "cfg_scale": cfg_scale,
    "temperature": temperature
}
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 Tensor, ndim 3, dtype family float, shape (2, 1, 8) — shape (2, 1, 8) 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_2d_sincos_pos_embed_from_grid Not yet run
unverified (harvested, no recorded run)
visual-gen/semanticist/semanticist/stage1/diffusion_transfomer.py
code served (permissive licence) · get_code("665d8a4e8f673a4c")
How it ran

no invocation recorded.

get_2d_sincos_pos_embed_from_grid Not yet run
unverified (harvested, no recorded run)
visual-gen/semanticist/semanticist/stage1/pos_embed.py
pointer only (licence: MIT); the stored copy get_code serves, from naver/croco, is pointer only (licence: NOASSERTION) · get_code("4fd80de79832745d")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from naver/croco, and GitHub could not classify the licence of that repository (NOASSERTION). 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. No invocation recorded.

instantiate_from_config Not yet run
unverified (harvested, no recorded run)
visual-gen/semanticist/semanticist/engine/trainer_utils.py
code served (permissive licence) · get_code("e6197485fe096b58")
How it ran

no invocation recorded.

Also present in other papers' code

Of 17 samples lifted from this paper's linked repositories, 17 carry a function identity and 0 do not. Of the 16 distinct functions with an identity, 10 are also present in other papers whose code we hold (9 with at least one paper dated earlier, 8 the same month, 10 later) and 6 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
get_2d_sincos_pos_embed_from_grid 190 other papers
146 dated earlier, 7 the same month, 34 later, 3 same year, order not determinable · 65 only as an assembly brick
1807.10221 (2018-07)
Unified Perceptual Parsing for Scene Understanding
2105.12723 (2021-05)
Nested Hierarchical Transformer: Towards Accurate, Data-Efficient and Interpretable Visual Understanding
2111.06377 (2021-11)
Masked Autoencoders Are Scalable Vision Learners
and 187 more not named here
modulate 101 other papers
68 dated earlier, 5 the same month, 26 later, 1 same year, order not determinable, 1 undated · 29 only as an assembly brick
2202.04200 (2022-02)
MaskGIT: Masked Generative Image Transformer
2203.10897 (2022-03)
Unified Multivariate Gaussian Mixture for Efficient Neural Image Compression
2212.09748 (2022-12)
Scalable Diffusion Models with Transformers
and 98 more not named here
get_2d_sincos_pos_embed 95 other papers
79 dated earlier, 5 the same month, 10 later, 1 same year, order not determinable · 33 only as an assembly brick
1807.10221 (2018-07)
Unified Perceptual Parsing for Scene Understanding
2105.12723 (2021-05)
Nested Hierarchical Transformer: Towards Accurate, Data-Efficient and Interpretable Visual Understanding
2111.06377 (2021-11)
Masked Autoencoders Are Scalable Vision Learners
and 92 more not named here
get_2d_sincos_pos_embed 69 other papers
49 dated earlier, 2 the same month, 18 later · 12 only as an assembly brick
2203.10897 (2022-03)
Unified Multivariate Gaussian Mixture for Efficient Neural Image Compression
2212.09748 (2022-12)
Scalable Diffusion Models with Transformers
2303.14389 (2023-03)
Masked Diffusion Transformer is a Strong Image Synthesizer
and 66 more not named here
get_obj_from_str 54 other papers
43 dated earlier, 2 the same month, 9 later · 12 only as an assembly brick
2006.11239 (2020-06)
Denoising Diffusion Probabilistic Models
2102.05379 (2021-02)
Argmax Flows and Multinomial Diffusion: Learning Categorical Distributions
2112.00384 (2021-12)
Exploration into Translation-Equivariant Image Quantization
and 51 more not named here
modulate 27 other papers
11 dated earlier, 3 the same month, 12 later, 1 undated · 2 only as an assembly brick
2312.15736 (2023-12)
Towards Real-World Blind Face Restoration with Generative Diffusion Prior
2404.11199 (2024-04)
RiboDiffusion: Tertiary Structure-based RNA Inverse Folding with Generative Diffusion Models
2405.17527 (2024-05)
Unisolver: PDE-Conditional Transformers Towards Universal Neural PDE Solvers
and 24 more not named here
find_multiple 13 other papers
8 dated earlier, 2 the same month, 2 later, 1 undated · 2 only as an assembly brick
2406.06525 (2024-06)
Autoregressive Model Beats Diffusion: Llama for Scalable Image Generation
2409.13689 (2024-09)
Temporally Aligned Audio for Video with Autoregression
2410.02705 (2024-10)
ControlAR: Controllable Image Generation with Autoregressive Models
and 10 more not named here
instantiate_from_config 13 other papers
10 dated earlier, 1 the same month, 2 later · 1 only as an assembly brick
2112.00384 (2021-12)
Exploration into Translation-Equivariant Image Quantization
2301.13670 (2023-01)
What Makes Good Examples for Visual In-Context Learning?
2306.14795 (2023-06)
MotionGPT: Human Motion as a Foreign Language
and 10 more not named here
build_mlp 4 other papers
1 dated earlier, 3 later · 1 only as an assembly brick
2412.08781 (2024-12)
GMem: A Modular Approach for Ultra-Efficient Generative Models
2508.10298 (2025-08)
SynBrain: Enhancing Visual-to-fMRI Synthesis via Probabilistic Representation Learning
2601.07773 (2026-01)
Self-transcendence: Is External Feature Guidance Indispensable for Accelerating Diffusion Transformer Training?
and 1 more not named here
drop_path 1 other paper
1 later
2506.22624 (2025-06)
Seg-R1: Segmentation Can Be Surprisingly Simple with Reinforcement Learning

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

https://visual-gen.github.io/semanticist Figure 1. Image reconstruction using our structured visual tokenization approach, which uniquely enables decoding at any token count. Each column shows reconstructions resulting from progressively increasing the number of tokens, from a single token to 256 tokens. Unlike conventional tokenizers that require a fixed number of tokens for meaningful decoding, our method ensures that each token incrementally refines the image, with earlier tokens capturing the most salient features and later ones adding finer details. This demonstrates the flexibility and effectiveness of our approach in producing coherent images even with very few tokens (view more from Fig. 17 in the Appendix).

For agents

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

get_harvested_code_for_paper("2503.08685")
get_code_for_paper("2503.08685")
have("2503.08685")

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

curl https://syntology.ai/api/ran/2503.08685.json

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

[![Syntology run record](https://syntology.ai/api/ran/2503.08685.svg)](https://syntology.ai/paper/2503.08685)

Connect an agent — have() is free.