SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2405.01719 · ICML · 2024

Inherent Trade-Offs between Diversity and Stability in Multi-Task Benchmarks

Moritz Hardt, Guanhua Zhang

arXiv · PDF · Open in the Atlas

Code that ran

9 of 12 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, 2 violated it and 0 have no contract verdict; 4 failures of our instrument; 3 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2405.01719.json · badge: /api/ran/2405.01719.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
socialfoundations/benchbench canonical (classifier label)
classifier basis: census
9 ran · 3 no recorded run
FunctionStatusWhere it lives
appr_rank_diff Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
socialfoundations/benchbench/benchbench/measures/cardinal.py
code served (permissive licence) · get_code("1100ae4be1bf2756")
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, shape (1,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Create sample score array - scores for models across tasks
score = np.array([0.5, 0.7, 0.6, 0.8], dtype=np.float32)

# Create sample old_rank array - original ranks
old_rank = np.array([3, 1, 2, 0], dtype=np.int32)

# Define use_weighted_loss parameter
use_weighted_loss = False

ARGS = {
    "score": score,
    "old_rank": old_rank,
    "use_weighted_loss": use_weighted_loss,
}
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, shape (1,) — shape (1,) 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
appr_rank_diff Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
socialfoundations/benchbench/benchbench/measures/ordinal.py
code served (permissive licence) · get_code("85be7ca4d40885ca")
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 numpy as np
import torch

torch.manual_seed(0)

# Construct arguments for appr_rank_diff
new_win_rate = np.array([0.5, 0.6, 0.7, 0.8, 0.9])
inv_indices = [0, 2, 4]
orig_rank = np.array([1.0, 3.0, 5.0])

ARGS = {
    "new_win_rate": new_win_rate,
    "inv_indices": inv_indices,
    "orig_rank": orig_rank,
}
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
get_combinations Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
socialfoundations/benchbench/benchbench/utils/base.py
code served (permissive licence) · get_code("d8821cf389d6c663")
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 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
from itertools import combinations

# Set seed for reproducibility
torch.manual_seed(0)

# Construct valid arguments for get_combinations function
# The function takes a list s and an integer k
# It returns all combinations of size k from the list s

ARGS = {
    "s": [1, 2, 3, 4, 5],
    "k": 2
}
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 list, ndim 2, 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_kendall_tau Ran
ran, honoured the contract we drafted
socialfoundations/benchbench/benchbench/utils/metric.py
code served (permissive licence) · get_code("baefae3daf59272f")
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 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 numpy as np
from scipy.stats import kendalltau

# Create two distinct ranking arrays
# Rankings should be arrays of integers or floats representing ranks/positions
new_rank = np.array([1, 2, 3, 4, 5])
old_rank = np.array([1, 3, 2, 5, 4])

ARGS = {
    "new_rank": new_rank,
    "old_rank": old_rank,
}
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 tuple, ndim 1, 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
get_kendall_w Ran
ran, honoured the contract we drafted
socialfoundations/benchbench/benchbench/utils/metric.py
code served (permissive licence) · get_code("9f7f918c7b987faa")
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 float64, ndim 0, 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 numpy as np

# The function expects rankings to be a list that can be converted to a numpy array
# Looking at the function:
# - rankings is converted to a numpy array with shape (m, n)
# - m is the number of rankings (axis 0)
# - n is the number of items being ranked (axis 1)
# - Each ranking contains rank values

# Create a simple example with 3 rankings of 4 items each
rankings = [
    [1, 2, 3, 4],
    [1, 3, 2, 4],
    [2, 1, 3, 4]
]

ARGS = {
    "rankings": rankings
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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_rank_diff Ran
ran, honoured the contract we drafted
socialfoundations/benchbench/benchbench/utils/metric.py
code served (permissive licence) · get_code("13665191512c83ca")
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 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 numpy as np

# Construct arguments for get_rank_diff
new_rank = np.array([2, 0, 3, 1])
old_rank = np.array([0, 1, 2, 3])

ARGS = {
    "new_rank": new_rank,
    "old_rank": old_rank
}
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 tuple, ndim 1, 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
get_selected_win_rate Ran
ran; our contract draft was wrong, not the code
socialfoundations/benchbench/benchbench/measures/ordinal.py
code served (permissive licence) · get_code("4837e4c35f8872e9")
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 numpy as np

torch.manual_seed(0)

# Create a small win rate matrix (4x4)
# Each row/column represents a model
# Entry [i,j] = win rate of model i over model j
win_rate_matrix = torch.tensor([
    [0.5, 0.6, 0.7, 0.8],
    [0.4, 0.5, 0.65, 0.75],
    [0.3, 0.35, 0.5, 0.7],
    [0.2, 0.25, 0.3, 0.5]
], dtype=torch.float32)

# Unnormalized log-probabilities for selecting each model (4 models)
w = torch.tensor([0.5, -0.3, 1.2, -0.8], dtype=torch.float32, requires_grad=True)

# Indices for L - which models are "invalid" initially
inv_indices = [1, 3]

# Flag for sampling vs deterministic selection
do_sample = False

ARGS = {
    "win_rate_matrix": win_rate_matrix,
    "w": w,
    "inv_indices": inv_indices,
    "do_sample": do_sample
}
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 — 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
is_int Ran
ran, violated the contract we drafted
socialfoundations/benchbench/benchbench/utils/base.py
code served (permissive licence) · get_code("706af7256ca3571e")
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 bool, 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

ARGS = {
    "x": "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 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
is_number Ran
ran, violated the contract we drafted
socialfoundations/benchbench/benchbench/utils/base.py
code served (permissive licence) · get_code("27f819180313d7a4")
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 bool, 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

torch.manual_seed(0)

# The function is_number takes a single parameter 's' which should be a string
# It attempts to convert the string to a float and returns True if successful,
# False if ValueError is raised

ARGS = {
    "s": "3.14"
}
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 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
get_diversity Not yet run
unverified (harvested, no recorded run)
socialfoundations/benchbench/benchbench/measures/cardinal.py
code served (permissive licence) · get_code("3a8ca465f7f286ad")
How it ran

no invocation recorded.

get_sensitivity Not yet run
unverified (harvested, no recorded run)
socialfoundations/benchbench/benchbench/measures/cardinal.py
code served (permissive licence) · get_code("2a977e4f402ff472")
How it ran

no invocation recorded.

get_sensitivity Not yet run
unverified (harvested, no recorded run)
socialfoundations/benchbench/benchbench/measures/ordinal.py
code served (permissive licence) · get_code("7a936e8599af8beb")
How it ran

no invocation recorded.

Also present in other papers' code

Of 12 samples lifted from this paper's linked repositories, 12 carry a function identity and 0 do not. Of the 12 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 1 later) and 11 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
is_number 13 other papers
7 dated earlier, 5 later, 1 undated
1202.3665 (2012-02)
emcee: The MCMC Hammer
1705.09980 (2017-05)
Neural Semantic Parsing by Character-based Translation: Experiments with Abstract Meaning Representations
2003.05855 (2020-03)
End-to-End Learning Local Multi-view Descriptors for 3D Point Clouds
and 10 more not named here

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 examine multi-task benchmarks in machine learning through the lens of social choice theory. We draw an analogy between benchmarks and electoral systems, where models are candidates and tasks are voters. This suggests a distinction between cardinal and ordinal benchmark systems. The former aggregate numerical scores into one model ranking; the latter aggregate rankings for each task. We apply Arrow's impossibility theorem to ordinal benchmarks to highlight the inherent limitations of ordinal systems, particularly their sensitivity to the inclusion of irrelevant models. Inspired by Arrow's theorem, we empirically demonstrate a strong trade-off between diversity and sensitivity to irrelevant changes in existing multi-task benchmarks. Our result is based on new quantitative measures of diversity and sensitivity that we introduce. Sensitivity quantifies the impact that irrelevant changes to tasks have on a benchmark. Diversity captures the degree of disagreement in model rankings across tasks. We develop efficient approximation algorithms for both measures, as exact computation is computationally challenging. Through extensive experiments on seven cardinal benchmarks and eleven ordinal benchmarks, we demonstrate a clear trade-off between diversity and stability: The more diverse a multi-task benchmark, the more sensitive to trivial changes it is. Additionally, we show that the aggregated rankings of existing benchmarks are highly unstable under irrelevant changes. The codes and data are available at https://socialfoundations. github.io/benchbench/.

For agents

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

get_harvested_code_for_paper("2405.01719")
get_code_for_paper("2405.01719")
have("2405.01719")

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

curl https://syntology.ai/api/ran/2405.01719.json

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

[![Syntology run record](https://syntology.ai/api/ran/2405.01719.svg)](https://syntology.ai/paper/2405.01719)

Connect an agent — have() is free.