SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2605.07950 · 2026

Slowly Annealed Langevin Dynamics: Theory and Applications to Training-Free Guided Generation

Yueming Lyu, Atsushi Nitanda, Dake Bu, Tanya Veeravalli

arXiv · PDF

Code that ran

13 of 15 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; 9 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 4 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/2605.07950.json · badge: /api/ran/2605.07950.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
anitan0925/sald canonical (classifier label)
classifier basis: census
13 ran · 2 no recorded run
FunctionStatusWhere it lives
alpha_tau_scalar Ran
ran, honoured the contract we drafted
anitan0925/sald/sald_Gaussian_Moon/sald_em_validation.py
code served (permissive licence) · get_code("f1a33b9aea7cd250")
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 float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import math
from dataclasses import dataclass

@dataclass(frozen=True)
class ModelConfig:
    T: float = 1.0
    beta_min: float = 0.1
    beta_max: float = 14.0
    x_mean: float = 1.0

def alpha_tau_scalar(tau: float, cfg: ModelConfig) -> float:
    integral_beta = cfg.beta_min * tau + 0.5 * (cfg.beta_max - cfg.beta_min) * tau * tau / cfg.T
    return math.exp(-0.5 * integral_beta)

# Create a ModelConfig instance
config = ModelConfig(
    T=1.0,
    beta_min=0.1,
    beta_max=14.0,
    x_mean=1.0
)

# Create a valid tau value
tau_value = 0.5

ARGS = {
    "tau": tau_value,
    "cfg": config
}
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 float, 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
beta_tau_scalar Ran
ran, honoured the contract we drafted
anitan0925/sald/sald_Gaussian_Moon/sald_eight_gaussian_validation.py
code served (permissive licence) · get_code("c309e75cdae04a4d")
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 float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
from dataclasses import dataclass

torch.manual_seed(0)

@dataclass(frozen=True)
class EightGaussianConfig:
    T: float
    beta_min: float
    beta_max: float
    centers: tuple[tuple[float, float], ...]
    penalty_mode_indices: tuple[int, ...]
    penalty_strength: float = 1.0
    penalty_width: float = 0.85

# Create a valid EightGaussianConfig instance
cfg = EightGaussianConfig(
    T=1.0,
    beta_min=0.1,
    beta_max=0.9,
    centers=((0.0, 0.0), (1.0, 1.0), (-1.0, 1.0), (1.0, -1.0)),
    penalty_mode_indices=(0, 1, 2, 3),
    penalty_strength=1.0,
    penalty_width=0.85
)

# Define arguments for beta_tau_scalar
ARGS = {
    "tau": 0.5,
    "cfg": cfg
}
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 float, 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
concat Ran
ran, honoured the contract we drafted
anitan0925/sald/sald_FM_SD35/utils.py
code served (permissive licence) · get_code("e585cbedd3948725")
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 (5, 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)

# Test case 1: List of torch.Tensor objects
tensor1 = torch.randn(2, 4)
tensor2 = torch.randn(3, 4)
data_tensors = [tensor1, tensor2]

# Test case 2: List of lists
data_lists = [[1, 2, 3], [4, 5, 6]]

# Use the tensor case for ARGS
ARGS = {
    "data": data_tensors
}
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 (5, 4) — shape (5, 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
configure_torch Ran
ran; our contract draft was wrong, not the code
anitan0925/sald/sald_Gaussian_Moon/sald_em_validation.py
code served (permissive licence) · get_code("6c78c68c3abc7e30")
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 device. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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

torch.manual_seed(0)

ARGS = {
    "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 device — a device output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a device 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
doit_reward_from_guidance Ran
ran, honoured the contract we drafted
anitan0925/sald/sald_Gaussian_Moon/sald_doit_validation.py
code served (permissive licence) · get_code("ab376366ff063880")
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)

# The function takes a guidance_value tensor and returns its negation
# guidance_value should be a torch.Tensor
guidance_value = torch.randn(2, 4)

ARGS = {
    "guidance_value": guidance_value,
}
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
get_image_transform Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
anitan0925/sald/sald_FM_SD35/rewards.py
code served (permissive licence) · get_code("315bbcfd462a106b")
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 Compose. 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
from transformers import AutoImageProcessor
import torchvision.transforms as T

torch.manual_seed(0)

class MockImageProcessor:
    def __init__(self):
        self.image_mean = [0.5, 0.5, 0.5]
        self.image_std = [0.2, 0.2, 0.2]
    
    def to_dict(self):
        return {
            'do_resize': True,
            'size': {'height': 224, 'width': 224},
            'do_center_crop': True,
            'crop_size': {'height': 224, 'width': 224},
            'do_normalize': True
        }

processor = MockImageProcessor()

ARGS = {
    'processor': processor
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, torchvision, transformers — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Compose — a Compose output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Compose 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_size Ran
ran, honoured the contract we drafted
anitan0925/sald/sald_FM_SD35/rewards.py
code served (permissive licence) · get_code("b4ae73baea6f9f47")
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 tuple, ndim 1, dtype family int; 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 torch
import numpy as np

torch.manual_seed(0)

# Analyze the function:
# - Takes a parameter 'size'
# - Can be an int: returns (size, size)
# - Can be a dict with 'height' and 'width': returns (height, width)
# - Can be a dict with 'shortest_edge': returns shortest_edge value
# - Otherwise raises ValueError

# We need to test one valid case. Let's use the int case as it's simplest.

ARGS = {
    'size': 256
}
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 tuple, ndim 1, dtype family int — 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
interpolate_guidance_potential Ran
ran; our contract draft was wrong, not the code
anitan0925/sald/sald_Gaussian_Moon/sald_guided_validation.py
code served (permissive licence) · get_code("a5ebf22cacb7b4df")
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 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 torch
import numpy as np
from dataclasses import dataclass

torch.manual_seed(0)

@dataclass(frozen=True)
class GuidedFieldConfig:
    lambda_guidance: float
    n_reference: int
    moon_noise: float
    moon_scale: float
    moon_shift_x: float
    moon_shift_y: float
    x_min: float
    x_max: float
    y_min: float
    y_max: float
    bins_x: int
    bins_y: int
    distance_eps: float = 0.001
    ref_chunk_size: int = 2048
    grid_chunk_size: int = 1024
    plot_reference_points: int = 12000

@dataclass
class GuidedFieldState:
    config: GuidedFieldConfig
    x_edges: torch.Tensor
    y_edges: torch.Tensor
    x_centers: torch.Tensor
    y_centers: torch.Tensor
    mesh_x: torch.Tensor
    mesh_y: torch.Tensor
    flat_points: torch.Tensor
    f_grid: torch.Tensor
    grad_grid: torch.Tensor
    f_field: torch.Tensor
    grad_field: torch.Tensor
    dx: float
    dy: float
    plot_reference_points: np.ndarray

# Create minimal GuidedFieldConfig
config = GuidedFieldConfig(
    lambda_guidance=1.0,
    n_reference=10,
    moon_noise=0.1,
    moon_scale=1.0,
    moon_shift_x=0.0,
    moon_shift_y=0.0,
    x_min=0.0,
    x_max=10.0,
    y_min=0.0,
    y_max=10.0,
    bins_x=4,
    bins_y=3
)

# Create minimal GuidedFieldState with required tensors
# f_field is used in _sample_field with grid_sample, needs shape (1, C, H, W)
x_edges = torch.linspace(0.0, 10.0, 5)
y_edges = torch.linspace(0.0, 10.0, 4)
x_centers = torch.linspace(0.5, 9.5, 4)
y_centers = torch.linspace(1.25, 8.75, 3)

mesh_x, mesh_y = torch.meshgrid(x_centers, y_centers, indexing='ij')
flat_points = torch.stack([mesh_x.flatten(), mesh_y.flatten()], dim=1)

# f_field: shape (1, 1, H, W) for grid_sample
f_field = torch.randn(1, 1, 3, 4)
grad_field = torch.randn(1, 2, 3, 4)
f_grid = torch.randn(4, 3)
grad_grid = torch.randn(4, 3, 2)

plot_reference_points_arr = np.random.randn(100, 2)

state = GuidedFieldState(
    config=config,
    x_edges=x_edges,
    y_edges=y_edges,
    x_centers=x_centers,
    y_centers=y_centers,
    mesh_x=mesh_x,
    mesh_y=mesh_y,
    flat_points=flat_points,
    f_grid=f_grid,
    grad_grid=grad_grid,
    f_field=f_field,
    grad_field=grad_field,
    dx=2.5,
    dy=3.33,
    plot_reference_points=plot_reference_points_arr
)

# Create points tensor (batch_size=2, 2D points)
points = torch.tensor([
    [2.0, 3.0],
    [7.5, 6.5]
], dtype=torch.float32)

ARGS = {
    "points": points,
    "state": state
}
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 Tensor, ndim 1, dtype family float — no output shape is stated: observe_signature.py could not observe it (import:TypeError)
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
make_circle_gaussian_centers Ran
ran, honoured the contract we drafted
anitan0925/sald/sald_Gaussian_Moon/sald_eight_gaussian_validation.py
code served (permissive licence) · get_code("4259d1b2ac33bcd5")
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 tuple, ndim 2, 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 math

ARGS = {
    "radius": 3.0,
    "start_angle": math.pi / 8.0,
}
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 tuple, ndim 2, 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
make_two_moons_samples Ran
ran, honoured the contract we drafted
anitan0925/sald/sald_Gaussian_Moon/sald_guided_validation.py
code served (permissive licence) · get_code("fbe8867da61cf978")
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 (10, 2) as a later run recorded it. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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)

ARGS = {
    'n_samples': 10,
    'noise': 0.1,
    'scale': 2.0,
    'shift_x': 1.5,
    'shift_y': -0.5,
    'seed': 42,
    'device': torch.device('cpu'),
}
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 (10, 2) — shape (10, 2) 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
mixture_radius_scalar Ran
ran, honoured the contract we drafted
anitan0925/sald/sald_Gaussian_Moon/sald_em_validation.py
code served (permissive licence) · get_code("e5036e68c9ac9c58")
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 float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import math
from dataclasses import dataclass

torch = None  # Not needed for this function

@dataclass(frozen=True)
class ModelConfig:
    T: float = 1.0
    beta_min: float = 0.1
    beta_max: float = 14.0
    x_mean: float = 1.0

def alpha_tau_scalar(tau: float, cfg: ModelConfig) -> float:
    integral_beta = cfg.beta_min * tau + 0.5 * (cfg.beta_max - cfg.beta_min) * tau * tau / cfg.T
    return math.exp(-0.5 * integral_beta)

def mixture_radius_scalar(t: float, cfg: ModelConfig) -> float:
    tau = max(0.0, min(cfg.T, cfg.T - t))
    return cfg.x_mean * alpha_tau_scalar(tau, cfg)

# Construct arguments
cfg = ModelConfig(
    T=1.0,
    beta_min=0.1,
    beta_max=14.0,
    x_mean=1.0
)

t = 0.5

ARGS = {
    "t": t,
    "cfg": cfg
}
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 float, 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
retry Ran
ran; our contract draft was wrong, not the code
anitan0925/sald/sald_FM_SD35/rewards.py
code served (permissive licence) · get_code("e3edbb3e2ded330b")
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 function. 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 functools
import time

def retry(times, failed_return, exceptions, backoff_factor=1):
    """A decorator for retrying a function upon specific exceptions."""

    def decorator(func):

        @functools.wraps(func)
        def wrapper(*args, **kwargs):
            attempt = 0
            while attempt < times:
                try:
                    return func(*args, **kwargs, retry_attempt=attempt)
                except exceptions as e:
                    print(f'Exception [{type(e)}:{e}] thrown when attempting to run {func}, attempt {attempt} of {times}')
                    time.sleep(backoff_factor * 2 ** attempt)
                    attempt += 1
            return failed_return
        return wrapper
    return decorator


# Setup code to construct valid arguments

# Define a simple function to decorate
def example_func(x, retry_attempt=0):
    """Simple function that may raise an exception."""
    if retry_attempt < 2:
        raise ValueError("Simulated error on attempt")
    return "success"

# Define the ARGS dictionary with all parameters
ARGS = {
    'times': 3,
    'failed_return': "failed_after_retries",
    'exceptions': (ValueError, RuntimeError),
    'backoff_factor': 0.5
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
reverse_beta_scalar Ran
ran, honoured the contract we drafted
anitan0925/sald/sald_Gaussian_Moon/sald_eight_gaussian_validation.py
code served (permissive licence) · get_code("4ce0caf207d3d34e")
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 float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np
from dataclasses import dataclass

torch.manual_seed(0)

@dataclass(frozen=True)
class EightGaussianConfig:
    T: float
    beta_min: float
    beta_max: float
    centers: tuple[tuple[float, float], ...]
    penalty_mode_indices: tuple[int, ...]
    penalty_strength: float = 1.0
    penalty_width: float = 0.85

# Create a valid EightGaussianConfig instance
cfg = EightGaussianConfig(
    T=1.0,
    beta_min=0.1,
    beta_max=0.9,
    centers=((0.0, 0.0), (1.0, 1.0), (-1.0, 0.0), (0.0, -1.0), 
             (1.0, -1.0), (-1.0, 1.0), (1.0, 0.5), (-1.0, -0.5)),
    penalty_mode_indices=(0, 1, 2, 3),
    penalty_strength=1.0,
    penalty_width=0.85
)

# reverse_beta_scalar takes t (float) and cfg (EightGaussianConfig)
t_value = 0.5

ARGS = {
    "t": t_value,
    "cfg": cfg
}
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 float, 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
batch_cache Not yet run
unverified (harvested, no recorded run)
anitan0925/sald/sald_FM_SD35/pipeline_extensions.py
code served (permissive licence) · get_code("d99f02fdf27310de")
How it ran

no invocation recorded.

prepare_guidance_field Not yet run
unverified (harvested, no recorded run)
anitan0925/sald/sald_Gaussian_Moon/sald_guided_validation.py
code served (permissive licence) · get_code("0776bf6cb364f0b3")
How it ran

no invocation recorded.

Also present in other papers' code

Of 15 samples lifted from this paper's linked repositories, 15 carry a function identity and 0 do not. Of the 15 distinct functions with an identity, 2 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 0 later) and 13 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_image_transform 3 other papers
3 dated earlier
2505.05470 (2025-05)
Flow-GRPO: Training Flow Matching Models via Online RL
2602.22809 (2026-02)
PhotoAgent: Exploratory Visual Aesthetic Planning with Large Vision Models
2603.18488 (2026-03)
TexEditor: Structure-Preserving Text-Driven Texture Editing
get_size 3 other papers
3 dated earlier
2505.05470 (2025-05)
Flow-GRPO: Training Flow Matching Models via Online RL
2602.22809 (2026-02)
PhotoAgent: Exploratory Visual Aesthetic Planning with Large Vision Models
2603.18488 (2026-03)
TexEditor: Structure-Preserving Text-Driven Texture Editing

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

We study Slowly Annealed Langevin Dynamics (SALD), a sampler for tracking a path of moving target distributions and approximating the terminal target through time slowdown. We establish non-asymptotic convergence guarantees via a KL differential inequality, showing that slowdown improves tracking through contraction of intermediate targets and the complexity of the path. Motivated by training-free guided generation with pretrained score-based generative models, we further introduce Velocity-Aware SALD (VA-SALD), which explicitly incorporates the underlying marginal distributions of the pretrained model and uses slowdown to correct the additional deviation induced by guidance. This yields a principled framework for trainingfree guided generation for diffusion-based and related generative model families, together with convergence guarantees that clarify the roles of intermediate functional inequalities and guidance bias. Code is available at https://github.com/anitan0925/sald.

For agents

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

get_harvested_code_for_paper("2605.07950")
get_code_for_paper("2605.07950")
have("2605.07950")

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

curl https://syntology.ai/api/ran/2605.07950.json

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

[![Syntology run record](https://syntology.ai/api/ran/2605.07950.svg)](https://syntology.ai/paper/2605.07950)

Connect an agent — have() is free.