SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2605.30675 · 2026

Human-Alignment, Calibration, and Activation Patterns in Large Language Model Uncertainty

Kyle Moore, Jesse Roberts, Daryl Watson, William Ward, Grayson Heyboer

arXiv · PDF · Open in the Atlas

Code that ran

7 of 9 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 3 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 4 failures of our instrument; 2 no recorded run), measured 2026-11-22; 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/2605.30675.json · badge: /api/ran/2605.30675.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
KyleAMoore/LLM-UQ-Align-and-Calibrate canonical (classifier label)
classifier basis: census
7 ran · 2 no recorded run
FunctionStatusWhere it lives
correlation Ran
ran, honoured the contract we drafted
KyleAMoore/LLM-UQ-Align-and-Calibrate/LinearProbing/10FoldLinearReg.py
code served (permissive licence) · get_code("242647e19db97dcf")
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
from scipy.stats import pearsonr

# The function takes two array-like parameters: y_true and y_pred
# Both should be 1D numeric arrays for correlation calculation

np.random.seed(0)

# Create simple 1D numpy arrays with some variation
y_true = np.array([1.0, 2.0, 3.0, 4.0, 5.0])
y_pred = np.array([1.1, 2.1, 2.9, 4.2, 4.8])

ARGS = {
    "y_true": y_true,
    "y_pred": y_pred,
}
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 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
find_unicode Ran
ran; our contract draft was wrong, not the code
KyleAMoore/LLM-UQ-Align-and-Calibrate/LinearProbing/ActivationGrabber.py
code served (permissive licence) · get_code("9c31e212f3db9283")
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 set. 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 pandas as pd
import re

# Create a minimal DataFrame that matches what find_unicode expects
# The function expects df['Question'] to be iterable strings
# and df['Answers'] to be iterable of iterables of strings

df = pd.DataFrame({
    'Question': [
        'What is the capital of France?',
        'How do you say hello in Spanish?'
    ],
    'Answers': [
        ['Paris', 'The capital city'],
        ['Hola', 'Buenos días']
    ]
})

ARGS = {
    'df': df
}
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 set — a set output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a set 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
flatten Ran
ran, honoured the contract we drafted
KyleAMoore/LLM-UQ-Align-and-Calibrate/Analysis/parquet_to_csv/parquet_to_csv.py
code served (permissive licence) · get_code("9594fce02d333ff9")
How it ran

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

import torch

torch.manual_seed(0)

# The function flatten takes a list of lists (lol) and flattens it into a single list
# It iterates through each sublist and adds each item to the result

# Create a simple list of lists with distinct sizes to be unambiguous
lol = [
    [1, 2, 3],
    [4, 5],
    [6, 7, 8, 9],
    [10]
]

ARGS = {
    "lol": lol
}
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, ndim 1, dtype family int — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
format_results Ran
ran; our contract draft was wrong, not the code
KyleAMoore/LLM-UQ-Align-and-Calibrate/LinearProbing/10FoldLinearReg.py
code served (permissive licence) · get_code("7f42302ce226f319")
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 numpy as np
import pandas as pd

# Create sample rows data
rows = [
    {
        'model': 'model_a',
        'layer': 'layer_1',
        'human_ent_mean_r2': 0.85,
        'human_ent_fold_r2s': [0.80, 0.90],
        'human_ent_mean_correlation': 0.92,
        'human_ent_fold_correlations': [0.90, 0.94],
        'human_ent_mean_mse': 0.05,
        'human_ent_fold_mses': [0.04, 0.06],
        'rt_correct_mean_r2': 0.78,
        'rt_correct_fold_r2s': [0.75, 0.81],
        'rt_correct_mean_correlation': 0.88,
        'rt_correct_fold_correlations': [0.86, 0.90],
        'rt_correct_mean_mse': 0.08,
        'rt_correct_fold_mses': [0.07, 0.09],
        'rt_combined_mean_r2': 0.81,
        'rt_combined_fold_r2s': [0.79, 0.83],
        'rt_combined_mean_correlation': 0.90,
        'rt_combined_fold_correlations': [0.88, 0.92],
        'rt_combined_mean_mse': 0.06,
        'rt_combined_fold_mses': [0.05, 0.07],
        'n_samples': 100,
        'n_features': 50,
    },
    {
        'model': 'model_a',
        'layer': 'layer_2',
        'human_ent_mean_r2': 0.87,
        'human_ent_fold_r2s': [0.82, 0.92],
        'human_ent_mean_correlation': 0.93,
        'human_ent_fold_correlations': [0.91, 0.95],
        'human_ent_mean_mse': 0.04,
        'human_ent_fold_mses': [0.03, 0.05],
        'rt_correct_mean_r2': 0.80,
        'rt_correct_fold_r2s': [0.77, 0.83],
        'rt_correct_mean_correlation': 0.89,
        'rt_correct_fold_correlations': [0.87, 0.91],
        'rt_correct_mean_mse': 0.07,
        'rt_correct_fold_mses': [0.06, 0.08],
        'rt_combined_mean_r2': 0.83,
        'rt_combined_fold_r2s': [0.81, 0.85],
        'rt_combined_mean_correlation': 0.91,
        'rt_combined_fold_correlations': [0.89, 0.93],
        'rt_combined_mean_mse': 0.05,
        'rt_combined_fold_mses': [0.04, 0.06],
        'n_samples': 120,
        'n_features': 55,
    },
    {
        'model': 'model_b',
        'layer': 'layer_1',
        'human_ent_mean_r2': 0.82,
        'human_ent_fold_r2s': [0.78, 0.86],
        'human_ent_mean_correlation': 0.90,
        'human_ent_fold_correlations': [0.88, 0.92],
        'n_samples': 110,
        'n_features': 52,
    },
]

ARGS = {
    'rows': rows,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type 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
get_corr Ran
ran, honoured the contract we drafted
KyleAMoore/LLM-UQ-Align-and-Calibrate/LinearProbing/10FoldRidgeLinearReg.py
code served (permissive licence) · get_code("127ad710b8b06315")
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
from scipy.stats import pearsonr

def get_corr(y_true, y_pred):
    if np.std(y_true) == 0 or np.std(y_pred) == 0:
        return np.nan
    try:
        corr, _ = pearsonr(y_true, y_pred)
        return float(corr)
    except Exception:
        return np.nan

# Set up arguments
np.random.seed(0)

# Create two arrays with some correlation
y_true = np.array([1.0, 2.0, 3.0, 4.0, 5.0])
y_pred = np.array([1.1, 2.2, 2.9, 4.1, 4.8])

ARGS = {
    "y_true": y_true,
    "y_pred": y_pred
}
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 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
run_probe Ran
ran; our contract draft was wrong, not the code
KyleAMoore/LLM-UQ-Align-and-Calibrate/LinearProbing/10FoldLinearReg.py
code served (permissive licence) · get_code("f503a0a114128c9d")
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 numpy as np
import pandas as pd

# Create a small sample dataframe with dim_ columns and a target column
np.random.seed(0)
n_samples = 20
n_features = 3

data = {}
for i in range(n_features):
    data[f'dim_{i}'] = np.random.randn(n_samples).astype(np.float32)

# Create a target column with some correlation to the features
target = (data['dim_0'] + 0.5 * data['dim_1'] + np.random.randn(n_samples) * 0.1).astype(np.float32)
data['target'] = target

df = pd.DataFrame(data)

ARGS = {
    'df': df,
    'target_col': 'target',
    'n_splits': 5,
    'random_state': 42
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas, scipy, sklearn — 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
run_probe Ran
ran; our contract draft was wrong, not the code
KyleAMoore/LLM-UQ-Align-and-Calibrate/LinearProbing/10FoldRidgeLinearReg.py
code served (permissive licence) · get_code("da86414c77a6c9e4")
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 numpy as np
import pandas as pd

# Create a small DataFrame with dim columns and a target column
np.random.seed(42)
n_samples = 20
n_features = 3

data = {}
for i in range(n_features):
    data[f'dim_{i}'] = np.random.randn(n_samples).astype(np.float32)

data['target'] = (
    data['dim_0'] * 0.5 + data['dim_1'] * 0.3 + data['dim_2'] * 0.2 + 
    np.random.randn(n_samples) * 0.1
).astype(np.float32)

df = pd.DataFrame(data)

ARGS = {
    'df': df,
    'target_col': 'target',
    'n_splits': 5,
    'random_state': 42
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas, scipy, sklearn — 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
load_model Not yet run
unverified (harvested, no recorded run)
KyleAMoore/LLM-UQ-Align-and-Calibrate/LinearProbing/ActivationGrabber.py
code served (permissive licence) · get_code("6012bb03063365be")
How it ran

no invocation recorded.

unicode_clean Not yet run
unverified (harvested, no recorded run)
KyleAMoore/LLM-UQ-Align-and-Calibrate/LinearProbing/ActivationGrabber.py
code served (permissive licence) · get_code("199710f300a9f7a1")
How it ran

no invocation recorded.

Also present in other papers' code

Of 9 samples lifted from this paper's linked repositories, 9 carry a function identity and 0 do not. Of the 9 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 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.

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

Uncertainty Quantification is a large and growing subfield of large language model behavioral analysis. Primarily to recognize and combat hallucination, the field has largely focused on measuring and improving calibration, the accuracy of uncertainty judgments to task efficacy. In this work, we investigate the relatively underexplored question of how similar large language model uncertainty is to human uncertainty. We investigate the presence and strength of human-similar uncertainty signals, deemed uncertainty alignment, in large language model overt behavior and internal activation patterns. We identify whether the models show evidence of simultaneous alignment and calibration on a variety of datasets covering both multiple choice and open ended factual recall. And we characterize the effect of instruct fine-tuning on each of these facets.

For agents

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

get_harvested_code_for_paper("2605.30675")
get_code_for_paper("2605.30675")
have("2605.30675")

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

curl https://syntology.ai/api/ran/2605.30675.json

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

[![Syntology run record](https://syntology.ai/api/ran/2605.30675.svg)](https://syntology.ai/paper/2605.30675)

Connect an agent — have() is free.