SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2402.11137 · NeurIPS · 2024

TuneTables: Context Optimization for Scalable Prior-Data Fitted Networks

Frank Hutter, Micah Goldblum, Valeriia Cherepanova, Robin Schirrmeister, Chinmay Hegde, Colin White, Niv Cohen, Benjamin Feuer

arXiv · PDF · Open in the Atlas

Code that ran

12 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; 0 honoured the contract we drafted, 1 violated it and 1 has no contract verdict; 10 failures of our instrument; 5 no recorded run); 1 of them was also run on the host, not in the jail, by the peer-agreement lane that set its level, 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/2402.11137.json · badge: /api/ran/2402.11137.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
penfever/TuneTables canonical (classifier label)
classifier basis: census
8 ran · 2 no recorded run
penfever/tunetables extension (classifier label)
classifier basis: classifier_2025
4 ran · 3 no recorded run
FunctionStatusWhere it lives
is_json_serializable Ran
ran, violated the contract we drafted
penfever/tunetables/tunetables/batch/run_tt_job.py
code served (permissive licence) · get_code("a5c44847d0bc55c2")
How it ran

Ran in Syntology's jail on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned bool, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. The comparison behind this sample's level is a different run: the peer-agreement lane ran it again on the host, not in the jail (report dated 2026-09-05T01:19:32Z). 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 json
import torch

torch.manual_seed(0)

# The function is_json_serializable tests if an object can be serialized to JSON
# It takes one parameter: obj (any type of object)
# We need to provide test cases that demonstrate both serializable and non-serializable objects

# Test with a JSON-serializable object (dict with basic types)
serializable_obj = {
    "name": "test",
    "value": 42,
    "nested": {"key": "value"},
    "list": [1, 2, 3],
    "float": 3.14,
    "bool": True,
    "null": None
}

ARGS = {
    "obj": serializable_obj
}
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 bool, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
JointBCELossWithLogits Ran
ran; our contract draft was wrong, not the code
penfever/TuneTables/tunetables/losses.py
code served (permissive licence) · get_code("890fb6ef88c9dbef")
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 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

torch.manual_seed(0)

# Analyzing the function:
# - output: tensor that gets unsqueezed and repeated
#   Shape progression: (batch, seq, hidden) -> (batch, seq, hidden, 1) -> (batch, seq, hidden, target_last_dim)
#   Then permuted to (hidden, batch, seq, target_last_dim)
# - target: tensor with shape[-1] used for repeat, and needs to broadcast with output after permute
#   Final shapes: output (hidden, batch, seq, target_last_dim), target must be compatible
#   After prod(-1): (hidden, batch, seq)
#   After mean(0): (batch, seq)
#   Result is scalar after final mean()

# Let's work backwards:
# - target.shape[-1] is used in repeat, let's say it's 5
# - output starts as (batch, seq, hidden), let's say (2, 3, 4)
# - After unsqueeze(-1): (2, 3, 4, 1)
# - After repeat(1, 1, 1, 5): (2, 3, 4, 5)
# - After permute(2, 0, 1, 3): (4, 2, 3, 5)
# - target needs to broadcast with (4, 2, 3, 5), so target could be (4, 2, 3, 5) or broadcastable

# For simplicity, let's make target the same shape or a compatible shape
batch_size = 2
seq_len = 3
hidden_dim = 4
target_last_dim = 5

output = torch.randn(batch_size, seq_len, hidden_dim)
target = torch.randint(0, 2, (hidden_dim, batch_size, seq_len, target_last_dim)).float()

ARGS = {
    "output": output,
    "target": target,
}
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 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
get_NormalInitializer Ran
ran; our contract draft was wrong, not the code
penfever/TuneTables/tunetables/initializers.py
code served (permissive licence) · get_code("4c84ce37831d7194")
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 torch
from torch import nn

torch.manual_seed(0)

# The function get_NormalInitializer takes a std parameter (standard deviation)
# and returns an initializer function that applies normal initialization to nn.Linear layers
# The std parameter should be a numeric value (float or int)

ARGS = {
    "std": 0.02
}
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_append_embedding_encoder Ran
ran; our contract draft was wrong, not the code
penfever/TuneTables/tunetables/encoders.py
code served (permissive licence) · get_code("939b599ed6041748")
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 torch
import torch.nn as nn

torch.manual_seed(0)

# Create a simple base encoder that takes (num_features) and (emsize) as arguments
class SimpleEncoder(nn.Module):
    def __init__(self, num_features, emsize):
        super().__init__()
        self.num_features = num_features
        self.emsize = emsize
        self.linear = nn.Linear(num_features, emsize)
    
    def forward(self, x):
        # x shape: (seq_len, batch_size, num_features)
        # output shape: (seq_len, batch_size, emsize)
        return self.linear(x)

# Create encoder_creator function that returns a SimpleEncoder
def encoder_creator(num_features, emsize):
    return SimpleEncoder(num_features, emsize)

ARGS = {
    "encoder_creator": encoder_creator
}
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_normalized_encoder Ran
ran; our contract draft was wrong, not the code
penfever/TuneTables/tunetables/encoders.py
code served (permissive licence) · get_code("39f5dc76953cf6db")
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 torch
import torch.nn as nn

torch.manual_seed(0)

# Define a simple encoder creator function
def simple_encoder_creator(in_dim, out_dim):
    return nn.Linear(in_dim, out_dim)

# Define data_std as a scalar float
data_std = 2.5

ARGS = {
    "encoder_creator": simple_encoder_creator,
    "data_std": data_std,
}
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_normalized_uniform_encoder Ran
ran; our contract draft was wrong, not the code
penfever/TuneTables/tunetables/encoders.py
code served (permissive licence) · get_code("16eab2bd7416886c")
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 torch
import torch.nn as nn
import math

torch.manual_seed(0)

# Define a minimal encoder_creator that satisfies the usage pattern
# The function expects encoder_creator to be callable with (in_dim, out_dim)
class SimpleLinearEncoder(nn.Module):
    def __init__(self, in_dim, out_dim):
        super().__init__()
        self.linear = nn.Linear(in_dim, out_dim)
    
    def forward(self, x):
        return self.linear(x)

# Create the encoder_creator function
def encoder_creator(in_dim, out_dim):
    return SimpleLinearEncoder(in_dim, out_dim)

ARGS = {
    "encoder_creator": encoder_creator
}
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_wandb_api_key Ran
ran; our contract draft was wrong, not the code
penfever/tunetables/tunetables/batch/run_tt_job.py
code served (permissive licence) · get_code("b91cdb6ff3f535f8")
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 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 os
import tempfile

# Create a temporary file with a test API key
temp_dir = tempfile.mkdtemp()
api_key_file = os.path.join(temp_dir, 'wandb_api_key.txt')
with open(api_key_file, 'w') as f:
    f.write('test_wandb_api_key_12345\n')

ARGS = {
    'api_key_file': api_key_file
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 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
kl_divergence Ran
ran; our contract draft was wrong, not the code
penfever/TuneTables/tunetables/losses.py
code served (permissive licence) · get_code("98d35d058dcc75c7")
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 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 torch.nn.functional as F

torch.manual_seed(0)

# Create two logit tensors with shape (batch_size=2, num_classes=4)
# These represent classifier outputs for 2 examples with 4 classes
clf_out_a = torch.randn(2, 4)
clf_out_b = torch.randn(2, 4)

ARGS = {
    'clf_out_a': clf_out_a,
    'clf_out_b': clf_out_b,
    'reduction': 'mean'
}
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 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
make_serializable Ran
ran; our contract draft was wrong, not the code
penfever/tunetables/tunetables/batch/run_tt_job.py
code served (permissive licence) · get_code("8022d5417a746548")
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 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 torch
import json

torch.manual_seed(0)

# Test cases for make_serializable function
ARGS = {
    "config_sample": {
        "tensor_field": torch.randn(2, 4),
        "nested_dict": {
            "value": 42,
            "tensor": torch.tensor([1.0, 2.0, 3.0]),
        },
        "list_field": [1, 2, torch.randn(2, 3)],
        "callable_field": lambda x: x + 1,
        "string_field": "hello",
        "int_field": 123,
        "float_field": 3.14,
        "bool_field": True,
        "none_field": None,
        "complex_number": complex(1, 2),
    }
}
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 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_compatibility_params Ran
ran; our contract draft was wrong, not the code
penfever/TuneTables/tunetables/train_loop.py
code served (permissive licence) · get_code("ab1f1da564c5ae70")
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 torch
import numpy as np
from argparse import Namespace

torch.manual_seed(0)

# Create a minimal config dict based on what the function accesses
config = {
    'recompute_attn': False,
    'prior_type': 'prior_bag',
    'differentiable': False,
    'flexible': False,
    'output_multiclass_ordered_p': 0.5,
    'differentiable_hyperparameters': {
        'output_multiclass_ordered_p': 0.5,
        'multiclass_type': 'ordinal',
        'sampling': 'uniform'
    },
    'multiclass_type': 'ordinal',
    'sampling': 'uniform',
    'pre_sample_causes': False,
    'multiclass_loss_type': 'cross_entropy',
    'categorical_feature_p': 0.1,
    'nan_prob_no_reason': 0.05,
    'nan_prob_unknown_reason': 0.05,
    'set_value_to_nan': 0.0,
    'new_mlp_per_example': False,
    'prior_mlp_scale_weights_sqrt': False,
    'batch_size_per_gp_sample': 32,
    'differentiable_hps_as_style': True,
    'normalize_ignore_label_too': True,
    'mix_activations': True,
    'balanced': True,
    'bptt': 100,
    'bptt_extra_samples': 10,
    'canonical_y_encoder': True,
    'total_available_time_in_s': 3600.0,
    'efficient_eval_masking': False
}

# Create a minimal args object with prior_type attribute
args = Namespace(
    prior_type='prior_bag'
)

ARGS = {
    'config': config,
    'args': args
}
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 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
taylor_linear_attention_forward Ran
ran; our contract draft was wrong, not the code
penfever/tunetables/tunetables/transformer.py
code served (permissive licence) · get_code("7646a622a583702c")
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 torch
import torch.nn as nn
from torch.nn.functional import dropout

torch.manual_seed(0)

# Set up small tensor dimensions
batch_size = 2
seq_len_tgt = 3
seq_len_src = 4
embed_dim = 8
num_heads = 2
head_dim = embed_dim // num_heads

# Create input tensors
query = torch.randn(seq_len_tgt, batch_size, embed_dim)
key = torch.randn(seq_len_src, batch_size, embed_dim)
value = torch.randn(seq_len_src, batch_size, embed_dim)

# Create projection weights and biases
in_proj_weight = torch.randn(3 * embed_dim, embed_dim)
in_proj_bias = torch.randn(3 * embed_dim)

# Output projection
out_proj_weight = torch.randn(embed_dim, embed_dim)
out_proj_bias = torch.randn(embed_dim)

# Optional tensors for bias and masking
bias_k = None
bias_v = None
key_padding_mask = None
attn_mask = None

# Static key/value for caching
static_k = None
static_v = None

# Separate projection weights (not used when use_separate_proj_weight=False)
q_proj_weight = None
k_proj_weight = None
v_proj_weight = None

ARGS = {
    'query': query,
    'key': key,
    'value': value,
    'embed_dim_to_check': embed_dim,
    'num_heads': num_heads,
    'in_proj_weight': in_proj_weight,
    'in_proj_bias': in_proj_bias,
    'bias_k': bias_k,
    'bias_v': bias_v,
    'add_zero_attn': False,
    'dropout_p': 0.0,
    'out_proj_weight': out_proj_weight,
    'out_proj_bias': out_proj_bias,
    'training': True,
    'key_padding_mask': key_padding_mask,
    'need_weights': True,
    'attn_mask': attn_mask,
    'use_separate_proj_weight': False,
    'q_proj_weight': q_proj_weight,
    'k_proj_weight': k_proj_weight,
    'v_proj_weight': v_proj_weight,
    'static_k': static_k,
    'static_v': static_v,
    'average_attn_weights': True,
    'is_causal': False,
    'taylor_order': 2,
}
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 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
taylor_linear_attention_forward Ran
ran on a synthesized input
penfever/TuneTables/tunetables/layer.py
code served (permissive licence) · get_code("8b70d1a53aec77a8")
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 torch
from torch import nn
import math

torch.manual_seed(0)

# Set up small tensor dimensions
batch_size = 2
seq_len_tgt = 3
seq_len_src = 4
embed_dim = 8
num_heads = 2
head_dim = embed_dim // num_heads

# Create input tensors
query = torch.randn(seq_len_tgt, batch_size, embed_dim)
key = torch.randn(seq_len_src, batch_size, embed_dim)
value = torch.randn(seq_len_src, batch_size, embed_dim)

# Create projection weights
in_proj_weight = torch.randn(3 * embed_dim, embed_dim)
in_proj_bias = torch.randn(3 * embed_dim)

# Output projection
out_proj_weight = torch.randn(embed_dim, embed_dim)
out_proj_bias = torch.randn(embed_dim)

# Optional parameters (not used in this case)
bias_k = None
bias_v = None
key_padding_mask = None
attn_mask = None
static_k = None
static_v = None
q_proj_weight = None
k_proj_weight = None
v_proj_weight = None

ARGS = {
    'query': query,
    'key': key,
    'value': value,
    'embed_dim_to_check': embed_dim,
    'num_heads': num_heads,
    'in_proj_weight': in_proj_weight,
    'in_proj_bias': in_proj_bias,
    'bias_k': bias_k,
    'bias_v': bias_v,
    'add_zero_attn': False,
    'dropout_p': 0.0,
    'out_proj_weight': out_proj_weight,
    'out_proj_bias': out_proj_bias,
    'training': True,
    'key_padding_mask': key_padding_mask,
    'need_weights': True,
    'attn_mask': attn_mask,
    'use_separate_proj_weight': False,
    'q_proj_weight': q_proj_weight,
    'k_proj_weight': k_proj_weight,
    'v_proj_weight': v_proj_weight,
    'static_k': static_k,
    'static_v': static_v,
    'average_attn_weights': True,
    'is_causal': False,
    'taylor_order': 2,
}
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 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
TaylorMultiheadAttention Not yet run
unverified (harvested, no recorded run)
penfever/tunetables/tunetables/transformer.py
code served (permissive licence) · get_code("79f5b9e40c967c9b")
How it ran

no invocation recorded.

TransformerEncoderLayer Not yet run
unverified (harvested, no recorded run)
penfever/tunetables/tunetables/transformer.py
code served (permissive licence) · get_code("3ca52e651c5bddce")
How it ran

no invocation recorded.

TransformerModel Not yet run
unverified (harvested, no recorded run)
penfever/tunetables/tunetables/transformer.py
code served (permissive licence) · get_code("3b298b0945e919bb")
How it ran

no invocation recorded.

eval_method Not yet run
unverified (harvested, no recorded run)
penfever/TuneTables/tunetables/run_baselines.py
code served (permissive licence) · get_code("9e7eb2a909a1fb2d")
How it ran

no invocation recorded.

reload_config Not yet run
unverified (harvested, no recorded run)
penfever/TuneTables/tunetables/train_loop.py
code served (permissive licence) · get_code("68a73e2ea3cbee4f")
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, 6 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 6 later) and 10 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
get_normalized_uniform_encoder 4 other papers
4 later
2403.10923 (2024-03)
Interpretable Machine Learning for TabPFN
2406.01805 (2024-06)
TabMDA: Tabular Manifold Data Augmentation for Any Classifier using Transformers with In-context Subsetting
2410.01565 (2024-10)
Bayes' Power for Explaining In-Context Learning Generalizations
and 1 more not named here
get_NormalInitializer 3 other papers
3 later
2403.10923 (2024-03)
Interpretable Machine Learning for TabPFN
2410.01565 (2024-10)
Bayes' Power for Explaining In-Context Learning Generalizations
2602.12082 (2026-02)
Empirical Gaussian Processes
get_normalized_encoder 3 other papers
3 later
2403.10923 (2024-03)
Interpretable Machine Learning for TabPFN
2406.01805 (2024-06)
TabMDA: Tabular Manifold Data Augmentation for Any Classifier using Transformers with In-context Subsetting
2410.01565 (2024-10)
Bayes' Power for Explaining In-Context Learning Generalizations
JointBCELossWithLogits 2 other papers
2 later
2403.10923 (2024-03)
Interpretable Machine Learning for TabPFN
2406.01805 (2024-06)
TabMDA: Tabular Manifold Data Augmentation for Any Classifier using Transformers with In-context Subsetting
get_append_embedding_encoder 2 other papers
2 later
2403.10923 (2024-03)
Interpretable Machine Learning for TabPFN
2406.01805 (2024-06)
TabMDA: Tabular Manifold Data Augmentation for Any Classifier using Transformers with In-context Subsetting
is_json_serializable 1 other paper
1 later
2405.19660 (2024-05)
PATIENT-Ψ: Using Large Language Models to Simulate Patients for Training Mental Health Professionals

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

While tabular classification has traditionally relied on from-scratch training, a recent breakthrough called prior-data fitted networks (PFNs) challenges this approach. Similar to large language models, PFNs make use of pretraining and in-context learning to achieve strong performance on new tasks in a single forward pass. However, current PFNs have limitations that prohibit their widespread adoption. Notably, TabPFN achieves very strong performance on small tabular datasets but is not designed to make predictions for datasets of size larger than 1000. In this work, we overcome these limitations and substantially improve the performance of PFNs via context optimization. We introduce TuneTables, a parameter-efficient fine-tuning strategy for PFNs that compresses large datasets into a smaller learned context. We conduct extensive experiments on nineteen algorithms over 98 datasets and find that TuneTables achieves the best performance on average, outperforming boosted trees such as CatBoost, while optimizing fewer than 5% of TabPFN's parameters. Furthermore, we show that TuneTables can be used as an interpretability tool and can even be used to mitigate biases by optimizing a fairness objective. We open-source our code and raw results at https://github.com/penfever/TuneTables.

For agents

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

get_harvested_code_for_paper("2402.11137")
get_code_for_paper("2402.11137")
have("2402.11137")

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

curl https://syntology.ai/api/ran/2402.11137.json

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

[![Syntology run record](https://syntology.ai/api/ran/2402.11137.svg)](https://syntology.ai/paper/2402.11137)

Connect an agent — have() is free.