SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2411.04424 · EMNLP · 2024

Bayesian Calibration of Win Rate Estimation with LLM Evaluators

Zhe Wang, Arman Cohan, Yicheng Gao, Gonghan Xu

arXiv · PDF · Open in the Atlas

Code that ran

4 of 4 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, 0 violated it and 0 have no contract verdict; 4 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/2411.04424.json · badge: /api/ran/2411.04424.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
yale-nlp/bay-calibration-llm-evaluators canonical (classifier label)
classifier basis: census
4 ran · 0 no recorded run
FunctionStatusWhere it lives
estimate_bernoulli_parameter Ran
ran; our contract draft was wrong, not the code
yale-nlp/bay-calibration-llm-evaluators/utils.py
code served (permissive licence) · get_code("77526c3df9498aa9")
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 rv_continuous_frozen. 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 scipy.stats import beta

torch.manual_seed(0)

# The function estimate_bernoulli_parameter takes:
# - s: number of successes (int or numeric)
# - n: number of trials (int or numeric)
# - prior_alpha: alpha parameter of beta prior (default 1)
# - prior_beta: beta parameter of beta prior (default 1)
# 
# The function calls beta(prior_alpha + s, prior_beta + n - s)
# and returns a scipy.stats.beta distribution object

s = 5  # number of successes
n = 10  # number of trials
prior_alpha = 2  # alpha parameter for beta prior
prior_beta = 2   # beta parameter for beta prior

ARGS = {
    "s": s,
    "n": n,
    "prior_alpha": prior_alpha,
    "prior_beta": prior_beta
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type rv_continuous_frozen — a rv_continuous_frozen output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a rv_continuous_frozen 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
extract_data_from_file Ran
ran; our contract draft was wrong, not the code
yale-nlp/bay-calibration-llm-evaluators/process_logs.py
code served (permissive licence) · get_code("343ed5023f5900d4")
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 DataFrame. 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 tempfile
import pathlib

# Create a temporary file with sample data matching the expected format
temp_file = tempfile.NamedTemporaryFile(mode='w', suffix='.txt', delete=False)
temp_file.write("""**************************
Comparing model_a and model_b
Difference between true p and estimated p mean: 0.123
Difference between true p and k: 0.456
phat by human label error: 0.789

**************************
Comparing model_c and model_d
Difference between true p and estimated p mean: 0.234
Difference between true p and k: 0.567
phat by human label error: 0.890

**************************
Comparing model_e and model_f
Difference between true p and estimated p mean: nan
Difference between true p and k: 0.678
phat by human label error: nan
""")
temp_file.close()

# Convert to Path object with stem attribute
file_path = pathlib.Path(temp_file.name)

ARGS = {
    'file_path': file_path,
    'human': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 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
plot_p Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
yale-nlp/bay-calibration-llm-evaluators/utils.py
code served (permissive licence) · get_code("09de802166b089fc")
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
import datetime

# Create sample p values for plotting
np.random.seed(42)
p_samples = np.random.beta(5, 2, size=10000)  # Generate enough samples to pass the 0.5 * sample_size check

ARGS = {
    'p': p_samples,
    'plot_title': 'Test P Distribution',
    'confidence': 0.95,
    'sample_size': 10000,
    'file_name': None,
    'true_p': 0.7,
    'k_as_p': 0.65,
    'freq_p': 0.68,
    'truth_mat_p': 0.72,
    'save_dir': '/dev/null'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
matplotlib, 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
sample_q Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
yale-nlp/bay-calibration-llm-evaluators/utils.py
code served (permissive licence) · get_code("7fe2951898a41c2d")
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, 100) 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.stats import beta

# For 'OneCoinBetaBernoulli' method
# q_dist_list should be a list of tuples (alpha, beta) for beta distribution
q_dist_list_one_coin = [
    (2.0, 5.0),
    (3.0, 4.0),
    (1.5, 2.5),
]

# For 'BetaBernoulli' method
# q_dist_list should be a 2D array with shape (n_coins, 2, 2)
# where each element is (alpha, beta) pair
q_dist_list_beta_bernoulli = np.array([
    [(2.0, 5.0), (3.0, 4.0)],
    [(1.5, 2.5), (2.5, 3.5)],
    [(1.0, 2.0), (3.0, 5.0)],
])

# Test with OneCoinBetaBernoulli
ARGS = {
    'q_method': 'OneCoinBetaBernoulli',
    'q_dist_list': q_dist_list_one_coin,
    'sample_size': 100,
}
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 2, dtype family float, shape (3, 100) — shape (3, 100) 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

Also present in other papers' code

Of 4 samples lifted from this paper's linked repositories, 4 carry a function identity and 0 do not. Of the 4 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 4 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.

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

Recent advances in large language models (LLMs) show the potential of using LLMs as evaluators for assessing the quality of text generations from LLMs. However, applying LLM evaluators naively to compare different systems can lead to unreliable results due to the inaccuracy and intrinsic bias of LLM evaluators. In order to mitigate this problem, we propose two calibration methods, Bayesian Win-Rate Sampling (BWRS) and Bayesian Dawid-Skene, both of which leverage Bayesian inference to more accurately infer the true win rate of generative language models. We empirically validate our methods on six datasets covering story generation, summarization, and instruction following tasks. We show that both our methods are effective in improving the accuracy of win rate estimation using LLMs as evaluators, offering a promising direction for reliable automatic text quality evaluation.

For agents

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

get_harvested_code_for_paper("2411.04424")
get_code_for_paper("2411.04424")
have("2411.04424")

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

curl https://syntology.ai/api/ran/2411.04424.json

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

[![Syntology run record](https://syntology.ai/api/ran/2411.04424.svg)](https://syntology.ai/paper/2411.04424)

Connect an agent — have() is free.