SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2311.01933 · 2023

ForecastPFN: Synthetically-Trained Zero-Shot Forecasting

arXiv · PDF · Open in the Atlas

Code that ran

8 of 8 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; 3 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 5 failures 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/2311.01933.json · badge: /api/ran/2311.01933.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
abacusai/forecastpfn canonical (classifier label)
classifier basis: census
8 ran · 0 no recorded run
FunctionStatusWhere it lives
build_input Ran
ran; our contract draft was wrong, not the code
abacusai/forecastpfn/src/build_validation_dataset.py
code served (permissive licence) · get_code("c606881bead7833f")
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. 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
import pandas as pd
import tensorflow as tf

# Set seeds for reproducibility
np.random.seed(0)
tf.random.set_seed(0)

# Define parameters for build_input function
HISTORY = 100
HORIZON = 10

# Create sample timestamp array - needs to be large enough to work with HISTORY and HORIZON
# Create 120 timestamps (more than HISTORY + HORIZON)
ts = pd.date_range('2023-01-01', periods=120, freq='D').values

# Create sample target values - array of 120 float values
target_full = np.random.randn(120).astype(np.float32)

# task parameter can be 1, 2, or 3
task = 1

# Construct ARGS dictionary
ARGS = {
    'ts': ts,
    'target_full': target_full,
    'task': task
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
compute_time_features Ran
ran, honoured the contract we drafted
abacusai/forecastpfn/src/build_validation_dataset.py
code served (permissive licence) · get_code("3f0ae68d827e3d97")
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 int, shape (2, 5) 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 pandas as pd
import numpy as np

# The function takes a ts parameter which is an np.ndarray
# It converts it to datetime and extracts time features
# We need to provide an array that can be converted to datetime

# Create a small array of datetime strings
ts = np.array(['2023-01-15', '2023-06-20'], dtype='datetime64[D]')

ARGS = {
    'ts': ts
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas — 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 int, shape (2, 5) — shape (2, 5) 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
decor_time Ran
ran; our contract draft was wrong, not the code
abacusai/forecastpfn/benchmark/layers/AutoCorrelation.py
code served (permissive licence) · get_code("c4d0804ed8acef6e")
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 time
import torch
import torch.nn as nn

def decor_time(func):
    def func2(*args, **kw):
        now = time.time()
        y = func(*args, **kw)
        t = time.time() - now
        print('call <{}>, time={}'.format(func.__name__, t))
        return y
    return func2

# Create a simple test function to decorate
@decor_time
def sample_function(x, y):
    """A simple function that does some work"""
    result = x + y
    time.sleep(0.001)  # Small delay to ensure measurable time
    return result

# Define ARGS mapping every parameter of decor_time to a value
ARGS = {
    'func': sample_function
}
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 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
get_frequency_modes Ran
ran, honoured the contract we drafted
abacusai/forecastpfn/benchmark/layers/FourierCorrelation.py
code served (permissive licence) · get_code("592ea8b254b006db")
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 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 numpy as np

np.random.seed(0)

ARGS = {
    'seq_len': 128,
    'modes': 64,
    'mode_select_method': 'random'
}
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 list, ndim 1, dtype family int — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_phi_psi Ran
ran; our contract draft was wrong, not the code
abacusai/forecastpfn/benchmark/layers/utils.py
code served (permissive licence) · get_code("f32036738135c8a8")
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. 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 functools import partial

# Construct valid arguments for get_phi_psi
ARGS = {
    'k': 3,
    'base': 'legendre'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sympy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
legendreDer Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
abacusai/forecastpfn/benchmark/layers/utils.py
code served (permissive licence) · get_code("4ef26472da51e3aa")
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 (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 numpy as np
from scipy.special import eval_legendre

# The function legendreDer takes two parameters:
# k: an integer representing the order
# x: a value or array on which to evaluate the Legendre polynomials

# Looking at the function:
# - k is used in range(k-1, -1, -2) so it should be an integer
# - x is passed to eval_legendre which accepts scalars or arrays
# - The function iterates through odd/even indices based on k

# Create small test values
k = 5  # integer order
x = np.array([0.0, 0.5, -0.5, 1.0])  # small array of evaluation points

ARGS = {
    "k": k,
    "x": x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — 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 (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
phi_ Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
abacusai/forecastpfn/benchmark/layers/utils.py
code served (permissive licence) · get_code("a54c8c5c47c6a8b8")
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 (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 numpy as np

# phi_ takes phi_c (polynomial coefficients), x (input array), and bounds lb, ub
# It creates a Polynomial from phi_c, evaluates it at x, and masks values outside [lb, ub]

# Create polynomial coefficients (small array)
phi_c = np.array([1.0, 2.0, 0.5])

# Create input array with values in and out of bounds
x = np.array([0.2, 0.5, 0.8, 1.5])

# Define bounds
lb = 0.0
ub = 1.0

ARGS = {
    "phi_c": phi_c,
    "x": x,
    "lb": lb,
    "ub": ub,
}
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 1, dtype family float, 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
read_timeseries_file Ran
ran, honoured the contract we drafted
abacusai/forecastpfn/src/build_validation_dataset.py
code served (permissive licence) · get_code("fa112f4ccfdfa642")
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 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 csv
import tempfile
import os

# Create a temporary CSV file with time series data
temp_dir = tempfile.mkdtemp()
temp_csv = os.path.join(temp_dir, "timeseries.csv")

with open(temp_csv, 'w', newline='') as f:
    writer = csv.writer(f)
    writer.writerow([1.0, 2.0, 3.0, 4.0])
    writer.writerow([5.0, 6.0, 7.0, 8.0])
    writer.writerow([9.0, 10.0, 11.0, 12.0])

ARGS = {
    "filename": temp_csv
}
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 2, 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

Also present in other papers' code

Of 8 samples lifted from this paper's linked repositories, 8 carry a function identity and 0 do not. Of the 8 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 3 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_frequency_modes 27 other papers
5 dated earlier, 22 later · 1 only as an assembly brick
2201.12740 (2022-01)
FEDformer: Frequency Enhanced Decomposed Transformer for Long-term Series Forecasting
2304.09184 (2023-04)
Frequency Enhanced Hybrid Attention Network for Sequential Recommendation
2305.13036 (2023-05)
Disentangling Structured Components: Towards Adaptive, Interpretable and Scalable Time Series Forecasting
and 24 more not named here
phi_ 24 other papers
6 dated earlier, 18 later · 1 only as an assembly brick
2201.12740 (2022-01)
FEDformer: Frequency Enhanced Decomposed Transformer for Long-term Series Forecasting
2205.08897 (2022-05)
FiLM: Frequency improved Legendre Memory Model for Long-term Time Series Forecasting
2304.05206 (2023-04)
The Capacity and Robustness Trade-off: Revisiting the Channel Independent Strategy for Multivariate Time Series Forecasting
and 21 more not named here
get_phi_psi 23 other papers
5 dated earlier, 18 later · 1 only as an assembly brick
2201.12740 (2022-01)
FEDformer: Frequency Enhanced Decomposed Transformer for Long-term Series Forecasting
2205.08897 (2022-05)
FiLM: Frequency improved Legendre Memory Model for Long-term Time Series Forecasting
2305.13036 (2023-05)
Disentangling Structured Components: Towards Adaptive, Interpretable and Scalable Time Series Forecasting
and 20 more not named here
legendreDer 23 other papers
5 dated earlier, 18 later · 1 only as an assembly brick
2201.12740 (2022-01)
FEDformer: Frequency Enhanced Decomposed Transformer for Long-term Series Forecasting
2205.08897 (2022-05)
FiLM: Frequency improved Legendre Memory Model for Long-term Time Series Forecasting
2305.13036 (2023-05)
Disentangling Structured Components: Towards Adaptive, Interpretable and Scalable Time Series Forecasting
and 20 more not named here
decor_time 3 other papers
2 dated earlier, 1 later
2205.08897 (2022-05)
FiLM: Frequency improved Legendre Memory Model for Long-term Time Series Forecasting
2310.14838 (2023-10)
Calibration of Time-Series Forecasting: Detecting and Adapting Context-Driven Distribution Shift
2410.13792 (2024-10)
Analyzing Deep Transformer Models for Time Series Forecasting via Manifold 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

The vast majority of time-series forecasting approaches require a substantial training dataset. However, many real-life forecasting applications have very little initial observations, sometimes just 40 or fewer. Thus, the applicability of most forecasting methods is restricted in data-sparse commercial applications. While there is recent work in the setting of very limited initial data (so-called `zero-shot' forecasting), its performance is inconsistent depending on the data used for pretraining. In this work, we take a different approach and devise ForecastPFN, the first zero-shot forecasting model trained purely on a novel synthetic data distribution. ForecastPFN is a prior-data fitted network, trained to approximate Bayesian inference, which can make predictions on a new time series dataset in a single forward pass. Through extensive experiments, we show that zero-shot predictions made by ForecastPFN are more accurate and faster compared to state-of-the-art forecasting methods, even when the other methods are allowed to train on hundreds of additional in-distribution data points.

For agents

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

get_harvested_code_for_paper("2311.01933")
get_code_for_paper("2311.01933")
have("2311.01933")

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

curl https://syntology.ai/api/ran/2311.01933.json

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

[![Syntology run record](https://syntology.ai/api/ran/2311.01933.svg)](https://syntology.ai/paper/2311.01933)

Connect an agent — have() is free.