SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2109.06096 · 2021

The Grammar-Learning Trajectories of Neural Language Models

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 4 have no contract verdict; 0 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/2109.06096.json · badge: /api/ran/2109.06096.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
borgr/ordert canonical (classifier label)
classifier basis: census
4 ran · 0 no recorded run
FunctionStatusWhere it lives
accuracy_from_file Ran
ran on a synthesized input
borgr/ordert/transformers/borgr_code/analyse_results.py
code served (permissive licence) · get_code("bec434ab039b3074")
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 tempfile
import os

# Create a temporary file with exactly 2000 lines (1000 pairs of good/bad losses)
temp_file = tempfile.NamedTemporaryFile(mode='w', delete=False, suffix='.txt')
temp_filename = temp_file.name

try:
    # Write 2000 lines alternating between good and bad losses
    # Pattern: good_loss (even line), bad_loss (odd line)
    for i in range(2000):
        if i % 2 == 0:
            # Good loss - write a lower value
            temp_file.write(f"{0.5}\n")
        else:
            # Bad loss - write a higher value to make it > good_loss
            temp_file.write(f"{0.8}\n")
    temp_file.close()
    
    ARGS = {
        'file': temp_filename
    }
finally:
    pass
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — 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
average_correlation Ran
ran on a synthesized input
borgr/ordert/transformers/borgr_code/analyse_results.py
code served (permissive licence) · get_code("26fd2d193c001284")
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 itertools
from scipy.stats import spearmanr, pearsonr

def average_correlation(orders, other_orders=None, pearson=True):
    corr = 0
    ranks = []
    if not pearson:
        unique_orders = []
        for order in orders:
            unique_order = []
            for item in order:
                if item not in unique_order:
                    unique_order.append(item)
            unique_orders.append(unique_order)
        orders = unique_orders
    for order in orders:
        if pearson:
            ranks.append(order)
        else:
            ranks.append([orders[0].index(item) for item in order])
    if other_orders is None:
        pairs = itertools.combinations(ranks, 2)
    else:
        pairs = itertools.product(orders, other_orders)
    for pair_num, (rank_a, rank_b) in enumerate(pairs):
        if pearson:
            corr += pearsonr(rank_a, rank_b)[0]
        else:
            corr += spearmanr(rank_a, rank_b)[0]
    corr = corr / (pair_num + 1)
    return corr


ARGS = {
    "orders": [
        [1.0, 2.0, 3.0, 4.0],
        [1.5, 2.5, 3.5, 4.5],
        [0.5, 1.5, 2.5, 3.5]
    ],
    "other_orders": None,
    "pearson": True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
scipy — 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
learnt_orders Ran
ran on a synthesized input
borgr/ordert/transformers/borgr_code/analyse_results.py
code served (permissive licence) · get_code("66d59c294b8e06d5")
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 pandas as pd
import numpy as np

# Create a sample DataFrame that matches what the function expects
data = {
    'model': ['model_a', 'model_a', 'model_a', 'model_b', 'model_b', 'model_b'],
    'challenge': ['challenge_1', 'challenge_2', 'challenge_3', 'challenge_1', 'challenge_2', 'challenge_3'],
    'steps': [10, 10, 10, 10, 10, 10],
    'accuracy': [0.85, 0.72, 0.91, 0.80, 0.75, 0.88]
}

df = pd.DataFrame(data)

# Create scores parameter - this should match a value in the 'steps' column
scores = 10

ARGS = {
    'df': df,
    'scores': scores,
    'measure': 'steps',
    'pearson': True
}
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 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
unigram_sentence_prob Ran
ran on a synthesized input
borgr/ordert/kenlm/blimp_eval.py
code served (permissive licence) · get_code("33bf519d8bdc6134")
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 torch

torch.manual_seed(0)

# The function expects:
# - log_probs: a dict mapping words to log probability values
# - words: a list/iterable of words
# - default_log_prob: optional float (default None)

log_probs = {
    "the": -0.5,
    "quick": -1.2,
    "brown": -1.5,
    "fox": -2.0,
    "jumps": -1.8,
}

words = ["the", "quick", "brown", "fox"]

ARGS = {
    "log_probs": log_probs,
    "words": words,
    "default_log_prob": None,
}
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 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

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

The learning trajectories of linguistic phenomena in humans provide insight into linguistic representation, beyond what can be gleaned from inspecting the behavior of an adult speaker. To apply a similar approach to analyze neural language models (NLM), it is first necessary to establish that different models are similar enough in the generalizations they make. In this paper, we show that NLMs with different initialization, architecture, and training data acquire linguistic phenomena in a similar order, despite their different end performance. These findings suggest that there is some mutual inductive bias that underlies these models' learning of linguistic phenomena. Taking inspiration from psycholinguistics, we argue that studying this inductive bias is an opportunity to study the linguistic representation implicit in NLMs. Leveraging these findings, we compare the relative performance on different phenomena at varying learning stages with simpler reference models. Results suggest that NLMs exhibit consistent "developmental" stages. Moreover, we find the learning trajectory to be approximately one-dimensional: given an NLM with a certain overall performance, it is possible to predict what linguistic generalizations it has already acquired. Initial analysis of these stages presents phenomena clusters (notably morphological ones), whose performance progresses in unison, suggesting a potential link between the generalizations behind them.

For agents

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

get_harvested_code_for_paper("2109.06096")
get_code_for_paper("2109.06096")
have("2109.06096")

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

curl https://syntology.ai/api/ran/2109.06096.json

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

[![Syntology run record](https://syntology.ai/api/ran/2109.06096.svg)](https://syntology.ai/paper/2109.06096)

Connect an agent — have() is free.