SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2303.09092 · 2023

Challenges to Evaluating the Generalization of Coreference Resolution Models: A Measurement Modeling Perspective

arXiv · PDF · Open in the Atlas

Code that ran

5 of 7 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 5 have no contract verdict; 0 failures of our instrument; 2 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/2303.09092.json · badge: /api/ran/2303.09092.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
ianporada/coref-data canonical (classifier label)
classifier basis not recorded on the link
5 ran · 2 no recorded run
FunctionStatusWhere it lives
compound_modifier Ran
ran on a synthesized input
ianporada/coref-data/analysis/types_of_coreference.py
code served (permissive licence) · get_code("5ff20a7cf05ae9ec")
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 = {
    "mention": (0, 0, 2),
    "sentences": [
        [
            ["word0", "pos0", "lemma0", "head0", "deprel0", "extra0", "extra1", "nn"],
            ["word1", "pos1", "lemma1", "head1", "deprel1", "extra1", "extra2", "compound"],
            ["word2", "pos2", "lemma2", "head2", "deprel2", "extra2", "extra3", "nn"],
        ],
        [
            ["word3", "pos3", "lemma3", "head3", "deprel3", "extra3", "extra4", "nsubj"],
            ["word4", "pos4", "lemma4", "head4", "deprel4", "extra4", "extra5", "compound"],
        ],
    ]
}
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
generic Ran
ran on a synthesized input
ianporada/coref-data/analysis/types_of_coreference.py
code served (permissive licence) · get_code("d0ec74767a7beadb")
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

# Build sentences structure
# Each sentence is a list of lines (word tokens)
# Each line is a list where:
#   TEXT_COL (1) = word text
#   XPOS_COL (4) = POS tag
#   HEAD_COL (6) = head index
#   DEPREL_COL (7) = dependency relation

# Create sample sentences with proper structure
def create_token_line(text, xpos, head, deprel):
    """Create a token line with required columns"""
    line = [None] * 8  # Need at least 8 columns (0-7)
    line[1] = text      # TEXT_COL
    line[4] = xpos      # XPOS_COL
    line[6] = str(head) # HEAD_COL
    line[7] = deprel    # DEPREL_COL
    return line

# Sentence 1: contains "a" with det relation
sentence1 = [
    create_token_line("The", "DT", 2, "det"),
    create_token_line("a", "DT", 3, "det"),
    create_token_line("cat", "NN", 0, "nsubj"),
    create_token_line("runs", "VBZ", 0, "root"),
]

# Sentence 2: contains determinant but head is NNS
sentence2 = [
    create_token_line("The", "DT", 2, "det"),
    create_token_line("dogs", "NNS", 0, "nsubj"),
    create_token_line("run", "VBP", 0, "root"),
]

# Sentence 3: no determinant, regular noun
sentence3 = [
    create_token_line("cat", "NN", 2, "nsubj"),
    create_token_line("runs", "VBZ", 0, "root"),
]

sentences = [sentence1, sentence2, sentence3]

# Create mentions (sent_id, start, end)
mention_with_a_det = (0, 0, 1)  # indices 0-1 in sentence1, includes "a" with det
mention_with_det_nns = (1, 0, 1)  # indices 0-1 in sentence2, has det, ends with NNS
mention_simple = (2, 0, 0)  # just "cat" in sentence3

ARGS = {
    "mention": mention_with_a_det,
    "sentences": sentences,
}
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
line_to_head_idx Ran
ran on a synthesized input
ianporada/coref-data/analysis/types_of_coreference.py
code served (permissive licence) · get_code("ad3a4129cea17a8b")
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 int, ndim 0, dtype family int, 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 line_to_head_idx takes a single parameter 'line'
# and accesses line[HEAD_COL] where HEAD_COL = 6
# It checks if line[HEAD_COL] is in ['-', '_', 'None'] and converts to int otherwise
# So 'line' should be a sequence (list, tuple, or array) with at least 7 elements (index 0-6)

# Create a valid line with at least 7 elements
# The element at index 6 should be convertible to int or be '-', '_', or 'None'
line = ['token1', 'token2', 'token3', 'token4', 'token5', 'token6', '3', 'extra']

ARGS = {
    'line': line
}
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 int, ndim 0, dtype family int, 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
parse_span Ran
ran on a synthesized input
ianporada/coref-data/preprocessing/indiscrim_conversion/corefud.py
code served (permissive licence) · get_code("6ce170087163c02d")
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 torch

ARGS = {
    'span': '1.5-3.7'
}
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 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
span_to_indices Ran
ran on a synthesized input
ianporada/coref-data/preprocessing/indiscrim_conversion/corefud.py
code served (permissive licence) · get_code("f0eb92137d72b363")
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

# Analyze the function:
# - span: tuple of (start, end) where start and end are 'id' values to match
# - tokens: list of dicts, each with 'id' and 'head' keys
# The function matches start/end ids to token positions and tracks min_head

# Create sample tokens with required 'id' and 'head' keys
tokens = [
    {'id': 1, 'head': None},
    {'id': 2, 'head': 5},
    {'id': 3, 'head': 2},
    {'id': 4, 'head': 3},
    {'id': 5, 'head': None},
]

# Create a span that will match valid token ids
span = (2, 4)  # start id=2 (index 1), end id=4 (index 3)

ARGS = {
    'span': span,
    'tokens': tokens,
}
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
convert_to_indiscrim Not yet run
unverified (harvested, no recorded run)
ianporada/coref-data/preprocessing/indiscrim_conversion/corefud.py
code served (permissive licence) · get_code("7f6dab293f9c000c")
How it ran

no invocation recorded.

get_examples Not yet run
unverified (harvested, no recorded run)
ianporada/coref-data/dataset_creation/manual/ezcoref_raw/push_to_hub.py
code served (permissive licence) · get_code("1eb237786c327189")
How it ran

no invocation recorded.

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 carry a function identity and 0 do not. Of the 7 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 7 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

It is increasingly common to evaluate the same coreference resolution (CR) model on multiple datasets. Do these multi-dataset evaluations allow us to draw meaningful conclusions about model generalization? Or, do they rather reflect the idiosyncrasies of a particular experimental setup (e.g., the specific datasets used)? To study this, we view evaluation through the lens of measurement modeling, a framework commonly used in the social sciences for analyzing the validity of measurements. By taking this perspective, we show how multi-dataset evaluations risk conflating different factors concerning what, precisely, is being measured. This in turn makes it difficult to draw more generalizable conclusions from these evaluations. For instance, we show that across seven datasets, measurements intended to reflect CR model generalization are often correlated with differences in both how coreference is defined and how it is operationalized; this limits our ability to draw conclusions regarding the ability of CR models to generalize across any singular dimension. We believe the measurement modeling framework provides the needed vocabulary for discussing challenges surrounding what is actually being measured by CR evaluations.

For agents

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

get_harvested_code_for_paper("2303.09092")
get_code_for_paper("2303.09092")
have("2303.09092")

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

curl https://syntology.ai/api/ran/2303.09092.json

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

[![Syntology run record](https://syntology.ai/api/ran/2303.09092.svg)](https://syntology.ai/paper/2303.09092)

Connect an agent — have() is free.