SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2410.13804 · 2024

BenTo: Benchmark Task Reduction with In-Context Transferability

arXiv · PDF · Open in the Atlas

Code that ran

12 of 13 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, 1 violated it and 0 have no contract verdict; 8 failures of our instrument; 1 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/2410.13804.json · badge: /api/ran/2410.13804.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
tianyi-lab/bento canonical (classifier label)
classifier basis: census
12 ran · 1 no recorded run
FunctionStatusWhere it lives
batch_infer_multiple_choice Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
tianyi-lab/bento/inference.py
code served (permissive licence) · get_code("d342730858d06a73")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. 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 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 torch.nn as nn

torch.manual_seed(0)

# Stand-in tokenizer class that mimics HuggingFace tokenizer behavior
class MockTokenizer:
    def batch_encode_plus(self, prompts, return_tensors=None, padding=False):
        # Return a dict with 'input_ids' and 'attention_mask' tensors
        batch_size = len(prompts)
        seq_len = 5
        input_ids = torch.randint(0, 1000, (batch_size, seq_len))
        attention_mask = torch.ones(batch_size, seq_len)
        return {
            'input_ids': input_ids,
            'attention_mask': attention_mask
        }
    
    def batch_decode(self, token_ids, skip_special_tokens=False):
        # Return a list of strings, one per batch item
        batch_size = token_ids.shape[0]
        return [f"The answer is A" for _ in range(batch_size)]

# Small stand-in model that has generate method
class MockModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 4)
    
    def generate(self, input_ids, attention_mask, max_new_tokens=1):
        # Generate method should return token ids
        batch_size = input_ids.shape[0]
        # Return batch_size x (seq_len + max_new_tokens) token ids
        output_seq_len = input_ids.shape[1] + max_new_tokens
        return torch.randint(0, 1000, (batch_size, output_seq_len))

# Create instances
tokenizer = MockTokenizer()
model = MockModel()

# Sample prompts for the function
prompts = [
    "What is the capital of France? A) Paris B) London C) Berlin D) Madrid",
    "What is 2+2? A) 3 B) 4 C) 5 D) 6",
    "Which is the largest planet? A) Earth B) Mars C) Jupiter D) Saturn",
    "What color is the sky? A) Green B) Red C) Blue D) Yellow"
]

ARGS = {
    'model': model,
    'tokenizer': tokenizer,
    'prompts': prompts
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, tqdm — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
batch_split Ran
ran; our contract draft was wrong, not the code
tianyi-lab/bento/inference.py
code served (permissive licence) · get_code("3bf1c3c00271897f")
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

torch.manual_seed(0)

# Define prompts as a list of strings
prompts = [
    "This is the first prompt.",
    "This is the second prompt.",
    "This is the third prompt.",
    "This is the fourth prompt.",
    "This is the fifth prompt.",
]

# Define batch_num as an integer
batch_num = 2

ARGS = {
    "prompts": prompts,
    "batch_num": batch_num,
}
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
close_cycle Ran
ran, violated the contract we drafted
tianyi-lab/bento/benchmark-reduction/plot.py
code served (permissive licence) · get_code("baa0ace81c394eb3")
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 torch

torch.manual_seed(0)

# close_cycle takes two parameters: r and theta
# Looking at the function body, it concatenates r[0] to r and theta[0] to theta
# This means r and theta should be lists (or list-like) where we can access the first element
# and use the + operator to concatenate

r = [1.0, 2.0, 3.0]
theta = [0.5, 1.0, 1.5]

ARGS = {
    "r": r,
    "theta": theta
}
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, 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
find_optimal_ordering Ran
ran; our contract draft was wrong, not the code
tianyi-lab/bento/benchmark-reduction/analysis.py
code served (permissive licence) · get_code("6fb31792767879ab")
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 sklearn.cluster import AgglomerativeClustering, KMeans
from scipy.sparse.csgraph import laplacian
from scipy.linalg import eigh

# Create a small similarity matrix (sparse-like but dense for simplicity)
# Using a 4x4 matrix for small size
np.random.seed(0)
n = 4
# Create a symmetric positive semi-definite matrix
A = np.random.randn(n, n)
similarity_matrix = A @ A.T + np.eye(n)

# Number of clusters
num_clusters = 2

# Matrix to rearrange (same size as similarity_matrix)
matrix_to_rearrange = np.random.randn(n, n)

ARGS = {
    "matrix": similarity_matrix,
    "num_clusters": num_clusters,
    "matrix_to_rearrange": matrix_to_rearrange,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy, sklearn — 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
format_example_flan Ran
ran; our contract draft was wrong, not the code
tianyi-lab/bento/prompts.py
code served (permissive licence) · get_code("be70f4753f03171a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

ARGS = {
    'data': {
        'inputs': 'What is the capital of France?',
        'targets': 'The capital of France is Paris.'
    },
    'include_answer': True,
    'include_question': 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 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
format_example_mmlu Ran
ran; our contract draft was wrong, not the code
tianyi-lab/bento/prompts.py
code served (permissive licence) · get_code("e068953fcd932df6")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

# Construct a valid data dict for the format_example_mmlu function
data = {
    'question': 'What is the capital of France?',
    'options': ['London', 'Berlin', 'Paris', 'Madrid'],
    'answer': 'C'
}

ARGS = {
    'data': data,
    'include_answer': True
}
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 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
gen_prompt_mmlu Ran
ran; our contract draft was wrong, not the code
tianyi-lab/bento/prompts.py
code served (permissive licence) · get_code("5cbdf932859e17b8")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import random
import numpy as np

class MockDataset:
    def __init__(self, data_list):
        self.data_list = data_list
        self.index = 0
    
    def shuffle(self):
        shuffled_data = self.data_list.copy()
        random.shuffle(shuffled_data)
        return MockDataset(shuffled_data)
    
    def __getitem__(self, idx):
        return self.data_list[idx]
    
    def __len__(self):
        return len(self.data_list)

trainset_data = [
    {
        'question': 'What is the capital of France?',
        'options': ['London', 'Berlin', 'Paris', 'Madrid'],
        'answer': 'C'
    },
    {
        'question': 'What is 2 + 2?',
        'options': ['3', '4', '5', '6'],
        'answer': 'B'
    },
    {
        'question': 'Which planet is closest to the Sun?',
        'options': ['Venus', 'Mercury', 'Earth', 'Mars'],
        'answer': 'B'
    },
    {
        'question': 'What is the largest ocean?',
        'options': ['Atlantic', 'Indian', 'Arctic', 'Pacific'],
        'answer': 'D'
    }
]

trainset = MockDataset(trainset_data)

ARGS = {
    'trainset': trainset,
    'example_num': 2,
    'eval_times': 1,
    'subject': 'high_school_biology'
}
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 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
load_csv Ran
ran, honoured the contract we drafted
tianyi-lab/bento/benchmark-reduction/analysis_compare_methods_.py
code served (permissive licence) · get_code("a4439ebb3e6536fd")
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 (3, 3) 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
import csv
import tempfile
import os

# Create a temporary CSV file for testing
temp_dir = tempfile.mkdtemp()
csv_path = os.path.join(temp_dir, 'test_benchmark.csv')

# Write a sample CSV file with header and data rows
with open(csv_path, 'w', newline='') as f:
    writer = csv.writer(f)
    writer.writerow(['metric1', 'metric2', 'metric3'])
    writer.writerow([1.5, 2.3, 3.7])
    writer.writerow([4.2, 5.1, 6.8])
    writer.writerow([7.9, 8.4, 9.2])

ARGS = {
    'path': csv_path
}
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 (3, 3) — shape (3, 3) 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
load_csv Ran
ran, honoured the contract we drafted
tianyi-lab/bento/benchmark-reduction/analysis_compare_methods_flan.py
code served (permissive licence) · get_code("1b7a6ef6b06e147c")
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 (3, 3) 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
import csv
import tempfile
import os

# Create a temporary CSV file with valid data
temp_dir = tempfile.mkdtemp()
temp_csv_path = os.path.join(temp_dir, 'test_data.csv')

# Write a valid CSV file with header and numeric data
with open(temp_csv_path, 'w', newline='') as f:
    writer = csv.writer(f)
    writer.writerow(['col1', 'col2', 'col3'])  # header row
    writer.writerow([1.0, 2.0, 3.0])
    writer.writerow([4.0, 5.0, 6.0])
    writer.writerow([7.0, 8.0, 9.0])

ARGS = {
    'path': temp_csv_path
}
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 (3, 3) — shape (3, 3) 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
load_csv Ran
ran, honoured the contract we drafted
tianyi-lab/bento/benchmark-reduction/analysis.py
code served (permissive licence) · get_code("5caec032cc3549d1")
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 (2, 3) 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
import tempfile
import os
import csv

# Create a temporary CSV file
temp_dir = tempfile.mkdtemp()
csv_path = os.path.join(temp_dir, "test_data.csv")

# Write a simple CSV with header and data
with open(csv_path, 'w', newline='') as f:
    writer = csv.writer(f)
    writer.writerow(['col1', 'col2', 'col3'])
    writer.writerow([1.0, 2.0, 3.0])
    writer.writerow([4.0, 5.0, 6.0])

ARGS = {
    'path': csv_path
}
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 (2, 3) — shape (2, 3) 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
prepare_input Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
tianyi-lab/bento/inference.py
code served (permissive licence) · get_code("5dc37098fd3e0e07")
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. 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)

class MockTokenizer:
    def batch_encode_plus(self, prompts, return_tensors=None, padding=None):
        batch_size = len(prompts)
        input_ids = torch.randint(0, 1000, (batch_size, 5))
        attention_mask = torch.ones((batch_size, 5), dtype=torch.long)
        return {
            'input_ids': input_ids,
            'attention_mask': attention_mask
        }

tokenizer = MockTokenizer()
prompts = ["Hello world", "How are you"]

ARGS = {
    'tokenizer': tokenizer,
    'prompts': prompts
}
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 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
separate_rows Ran
ran; our contract draft was wrong, not the code
tianyi-lab/bento/benchmark-reduction/analysis_compare_methods_.py
code served (permissive licence) · get_code("96537f07b172960c")
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 3, 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

# Create a small numpy array with distinct dimensions (4 rows, 3 columns)
arr = np.array([
    [1, 2, 3],
    [4, 5, 6],
    [7, 8, 9],
    [10, 11, 12]
])

# Create indices to separate
indices = [0, 2]

ARGS = {
    "arr": arr,
    "indices": indices
}
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 tuple, ndim 3, 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
find_optimal_ordering Not yet run
unverified (harvested, no recorded run)
tianyi-lab/bento/benchmark-reduction/analysis_compare_methods_.py
code served (permissive licence) · get_code("e1e528295e3ea8a4")
How it ran

no invocation recorded.

Also present in other papers' code

Of 13 samples lifted from this paper's linked repositories, 13 carry a function identity and 0 do not. Of the 11 distinct functions with an identity, 2 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 2 later) and 9 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
batch_split 2 other papers
1 dated earlier, 1 later
2403.00758 (2024-03)
Mitigating Reversal Curse in Large Language Models via Semantic-aware Permutation Training
2411.10606 (2024-11)
AmoebaLLM: Constructing Any-Shape Large Language Models for Efficient and Instant Deployment
prepare_input 1 other paper
1 later
2411.10606 (2024-11)
AmoebaLLM: Constructing Any-Shape Large Language Models for Efficient and Instant Deployment

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

Evaluating large language models (LLMs) is costly: it requires the generation and examination of LLM outputs on a large-scale benchmark of various tasks. This paper investigates how to efficiently reduce the tasks used to benchmark LLMs without affecting the evaluation quality. Our study reveals that task transferability and relevance provide critical information to identify the most representative subset of tasks via optimizing a facility location function. We propose a practically efficient metric for estimating the transferability between two tasks via in-context learning (ICL). By analyzing the pairwise transferability, we can reduce tasks in a modern LLM benchmark (e.g., MMLU or FLAN) to 5% while inducing only a <4% difference to the evaluation on the original benchmark. Compared to prior works, our method is training-free, gradient-free, and highly efficient requiring ICL only.

For agents

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

get_harvested_code_for_paper("2410.13804")
get_code_for_paper("2410.13804")
have("2410.13804")

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

curl https://syntology.ai/api/ran/2410.13804.json

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

[![Syntology run record](https://syntology.ai/api/ran/2410.13804.svg)](https://syntology.ai/paper/2410.13804)

Connect an agent — have() is free.