SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2408.05326 · 2024

A Psychology-based Unified Dynamic Framework for Curriculum Learning

arXiv · PDF · Open in the Atlas

Code that ran

6 of 6 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; 6 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/2408.05326.json · badge: /api/ran/2408.05326.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
nd-ball/cl-irt canonical (classifier label)
classifier basis: classifier_2025
6 ran · 0 no recorded run
FunctionStatusWhere it lives
calculate_theta_irt Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
nd-ball/cl-irt/PUDF_MedQA/deberta_pudf_medqa.py
code served (permissive licence) · get_code("2d2f163a314f52e0")
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) 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 norm

# Set random seed for reproducibility
np.random.seed(0)

# Create valid arguments for calculate_theta_irt
difficulties = [0.5, 1.2, -0.3, 0.8]
response_pattern = [1, -1, 1, -1]
num_obs = -1
initial_theta_val = 0.0

ARGS = {
    "difficulties": difficulties,
    "response_pattern": response_pattern,
    "num_obs": num_obs,
    "initial_theta_val": initial_theta_val,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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
create_dataset Ran
ran; our contract draft was wrong, not the code
nd-ball/cl-irt/PUDF_GLUE/PUDF_glue_DebertaV3.py
code served (permissive licence) · get_code("e6dd3a3e407b16d8")
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 TensorDataset. 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 torch.utils.data import TensorDataset

torch.manual_seed(0)

# Create a mock tokenizer that behaves like a HuggingFace tokenizer
class MockTokenizer:
    def __call__(self, text=None, text_pair=None, padding=None, truncation=None, max_length=None):
        # Return a dict with input_ids, attention_mask, and token_type_ids
        batch_size = len(text) if isinstance(text, list) else 1
        return {
            'input_ids': torch.randint(0, 1000, (batch_size, max_length or 32)),
            'attention_mask': torch.ones(batch_size, max_length or 32, dtype=torch.long),
            'token_type_ids': torch.zeros(batch_size, max_length or 32, dtype=torch.long)
        }

# Create a mock HuggingFace dataset
class MockHFDataset:
    def __init__(self, task):
        self.task = task
        self.data = self._create_data(task)
        self.column_names = list(self.data.keys())
        self.features = {col: None for col in self.column_names}
    
    def _create_data(self, task):
        data = {'label': [0, 1]}
        data['difficulty'] = [0.5, 0.7]
        
        if task == 'mnli':
            data['premise'] = ['This is a premise.', 'Another premise here.']
            data['hypothesis'] = ['This is a hypothesis.', 'Another hypothesis.']
        elif task in ['mrpc', 'rte']:
            data['sentence1'] = ['First sentence.', 'Another sentence.']
            data['sentence2'] = ['Second sentence.', 'Yet another.']
        elif task == 'qnli':
            data['question'] = ['What is this?', 'Who are you?']
            data['sentence'] = ['This is a sentence.', 'I am someone.']
        elif task == 'qqp':
            data['question1'] = ['First question?', 'Second question?']
            data['question2'] = ['Different question?', 'Another question?']
        elif task == 'sst2':
            data['sentence'] = ['Positive sentence here.', 'Negative sentence here.']
        
        return data
    
    def __getitem__(self, idx):
        result = {}
        for key, values in self.data.items():
            result[key] = values[idx]
        return result
    
    def map(self, func, batched=True, remove_columns=None, desc=None):
        # Apply function to all data at once
        batch_data = {k: v for k, v in self.data.items()}
        result_data = func(batch_data)
        
        # Keep only columns not in remove_columns
        if remove_columns:
            for col in remove_columns:
                if col in batch_data:
                    del batch_data[col]
        
        # Add tokenized columns to batch_data
        batch_data.update(result_data)
        
        return MockMappedDataset(batch_data, self.column_names)
    
    def set_format(self, type=None, columns=None):
        pass

class MockMappedDataset:
    def __init__(self, data, original_columns):
        self.data = data
        self.column_names = list(data.keys())
        self._columns_to_return = None
    
    def set_format(self, type=None, columns=None):
        self._columns_to_return = columns
    
    def __getitem__(self, idx):
        if isinstance(idx, str):
            # Return entire column
            return torch.tensor(self.data[idx]) if isinstance(self.data[idx], list) else self.data[idx]
        return self.data[idx]

# Create mock HF dataset for task 'mnli'
task = 'mnli'
hf_dataset = MockHFDataset(task)
tokenizer = MockTokenizer()
include_difficulty = True
partition_name = 'train'

ARGS = {
    'hf_dataset': hf_dataset,
    'task': task,
    'tokenizer': tokenizer,
    'include_difficulty': include_difficulty,
    'partition_name': partition_name
}
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 TensorDataset — a TensorDataset 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
select_data_for_pudf_epoch Ran
ran; our contract draft was wrong, not the code
nd-ball/cl-irt/PUDF_MedQA/deberta_pudf_medqa.py
code served (permissive licence) · get_code("12a39a39b0188061")
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 Dataset. 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 datasets import Dataset

torch.manual_seed(0)

# Create a minimal HuggingFace Dataset with difficulty column
data = {
    'text': ['sample1', 'sample2', 'sample3', 'sample4', 'sample5'],
    'difficulty': [0.1, 0.5, 0.8, 1.2, 1.5]
}
full_hf_train_dataset = Dataset.from_dict(data)

ARGS = {
    'full_hf_train_dataset': full_hf_train_dataset,
    'capacity_theta': 0.7,
    'difficulty_col': 'difficulty',
    'pudf_ordering': 'easiest',
    'lower_offset': -float('inf'),
    'upper_offset': 0.0,
    'min_samples_per_epoch': 100
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
datasets, numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Dataset — a Dataset output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Dataset output the recorded ndim 1 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
theta_fn_irt Ran
ran; our contract draft was wrong, not the code
nd-ball/cl-irt/PUDF_MedQA/deberta_pudf_medqa.py
code served (permissive licence) · get_code("4d0f6a7748882b67")
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) and returned function. 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
from scipy.special import expit
from scipy.stats import norm

# Create difficulties array - item difficulty parameters
difficulties = np.array([0.5, -0.3, 1.2, 0.1])

# Create student_prior - a scipy distribution object with logpdf method
student_prior = norm(loc=0, scale=1)

# Create response_pattern - array of 1 (correct), -1 (incorrect)
response_pattern = np.array([1, -1, 1, 1])

ARGS = {
    "difficulties": difficulties,
    "student_prior": student_prior,
    "response_pattern": response_pattern,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
tokenize_function Ran
ran; our contract draft was wrong, not the code
nd-ball/cl-irt/PUDF_GLUE/PUDF_glue_DebertaV3.py
code served (permissive licence) · get_code("3dc66ac0e9e83724")
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) 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
from transformers import AutoTokenizer

torch.manual_seed(0)

# Create a minimal tokenizer stand-in that mimics HuggingFace tokenizer behavior
class MockTokenizer:
    def __call__(self, text=None, text_pair=None, padding=None, truncation=None, max_length=None):
        # Return a dict with input_ids and attention_mask
        # Simulate tokenization by creating dummy token IDs
        if isinstance(text, list):
            batch_size = len(text)
        else:
            batch_size = 1
        
        # Create dummy token IDs of shape (batch_size, max_length)
        input_ids = torch.randint(0, 1000, (batch_size, max_length or 128))
        attention_mask = torch.ones((batch_size, max_length or 128), dtype=torch.long)
        
        return {
            'input_ids': input_ids,
            'attention_mask': attention_mask,
            'token_type_ids': torch.zeros((batch_size, max_length or 128), dtype=torch.long)
        }

# Create example data for different tasks
examples_mnli = {
    'premise': ['This is a premise.', 'Another premise here.'],
    'hypothesis': ['This is a hypothesis.', 'Another hypothesis.']
}

examples_mrpc = {
    'sentence1': ['Sentence one.', 'Another sentence.'],
    'sentence2': ['Sentence two.', 'Yet another.']
}

examples_rte = {
    'sentence1': ['Text A.', 'Text B.'],
    'sentence2': ['Text C.', 'Text D.']
}

examples_qqp = {
    'question1': ['What is this?', 'How does that work?'],
    'question2': ['What is that?', 'Why does it work?']
}

examples_qnli = {
    'question': ['What is the answer?', 'Where is it?'],
    'sentence': ['The answer is here.', 'It is over there.']
}

examples_sst2 = {
    'sentence': ['This movie is great!', 'I did not like it.']
}

# Create tokenizer
tokenizer = MockTokenizer()

# Create ARGS dict with all parameters
ARGS = {
    'examples': examples_mnli,
    'task': 'mnli',
    'tokenizer': tokenizer
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch, transformers — 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
tokenize_function Ran
ran; our contract draft was wrong, not the code
nd-ball/cl-irt/gen_difficulty/GLUE_difficulty.py
code served (permissive licence) · get_code("c19241ab56039094")
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) 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
from transformers import AutoTokenizer

torch.manual_seed(0)

# Create a minimal tokenizer stand-in that mimics HuggingFace tokenizer behavior
class MinimalTokenizer:
    def __call__(self, text1=None, text2=None, padding=None, truncation=None, max_length=None):
        # Simulate tokenizer output with input_ids and attention_mask
        if text2 is not None:
            # Two-text case
            batch_size = len(text1) if isinstance(text1, list) else 1
        else:
            # Single-text case
            batch_size = len(text1) if isinstance(text1, list) else 1
        
        return {
            'input_ids': torch.randint(0, 1000, (batch_size, max_length)),
            'attention_mask': torch.ones(batch_size, max_length)
        }

# Create example data for different tasks
examples_mnli = {
    'premise': ['This is a premise.', 'Another premise here.'],
    'hypothesis': ['This is a hypothesis.', 'Another hypothesis.']
}

examples_mrpc = {
    'sentence1': ['First sentence.', 'Another first.'],
    'sentence2': ['Second sentence.', 'Another second.']
}

examples_rte = {
    'sentence1': ['Text one.', 'Text two.'],
    'sentence2': ['Text three.', 'Text four.']
}

examples_qnli = {
    'question': ['What is this?', 'Why is that?'],
    'sentence': ['This is an answer.', 'That is an answer.']
}

examples_qqp = {
    'question1': ['Question A?', 'Question B?'],
    'question2': ['Question C?', 'Question D?']
}

examples_sst2 = {
    'sentence': ['This movie is great.', 'This movie is bad.']
}

tokenizer = MinimalTokenizer()

# Build ARGS dict with all parameters
ARGS = {
    'examples': examples_sst2,
    'task': 'sst2',
    'tokenizer': tokenizer
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch, transformers — 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

Also present in other papers' code

Of 6 samples lifted from this paper's linked repositories, 6 carry a function identity and 0 do not. Of the 6 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 6 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

Directly learning from examples of varying difficulty levels is often challenging for both humans and machine learning models. A more effective strategy involves exposing learners to examples in a progressive order from easy to difficult. Curriculum Learning (CL) has been proposed to implement this strategy in machine learning model training. However, two key challenges persist in CL framework design: defining the difficulty of training data and determining the appropriate amount of data to input at each training step. Drawing inspiration from psychometrics, this paper presents a Psychology-based Unified Dynamic Framework for Curriculum Learning (PUDF). We quantify the difficulty of training data by applying Item Response Theory (IRT) to responses from Artificial Crowds (AC). This theory-driven IRT-AC approach leads to global (i.e., model-independent) and interpretable difficulty values. Leveraging IRT, we propose a training strategy, Dynamic Data Selection via Model Ability Estimation (DDS-MAE), to schedule the appropriate amount of data during model training. Since our difficulty labeling and model ability estimation are based on a consistent theory, namely IRT, their values are comparable within the same scope, potentially leading to aligned training data selection and faster convergence compared to the other CL methods. Experimental results demonstrate that fine-tuning pre-trained large language models with PUDF leads to higher accuracy and faster convergence on a suite of benchmark datasets compared to standard fine-tuning and state-of-the-art CL methods. Ablation studies and downstream analyses further validate the impact of PUDF for CL.

For agents

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

get_harvested_code_for_paper("2408.05326")
get_code_for_paper("2408.05326")
have("2408.05326")

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

curl https://syntology.ai/api/ran/2408.05326.json

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

[![Syntology run record](https://syntology.ai/api/ran/2408.05326.svg)](https://syntology.ai/paper/2408.05326)

Connect an agent — have() is free.