SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2606.22305 · 2026

Learning at the Right Pace: Adaptive Data Scheduling Improves LLM Reinforcement Learning

Vladimir Braverman, Le, Guanchu Wang, Oren Gal, Zhaozhuo Xu, Anh Hoang, Ruixuan Zhang, Alexander Szalay, Zicheng Xu, Xiuyi Lou, Yu-Neng Chuang, Preprint

arXiv · PDF · Open in the Atlas

Code that ran

17 of 17 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; 1 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 16 failures of our instrument; 0 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/2606.22305.json · badge: /api/ran/2606.22305.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
Richard-zrx/ADS canonical (classifier label)
classifier basis: census
17 ran · 0 no recorded run
FunctionStatusWhere it lives
apply_overrides Ran
ran; our contract draft was wrong, not the code
Richard-zrx/ADS/src/cluster_embeddings.py
code served (permissive licence) · get_code("189c9ff78bd2b468")
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 argparse
import torch

# Create a base config dictionary
base_config = {
    'ids_path': '/default/ids/path',
    'embedding_path': '/default/embedding/path',
    'output_dir': '/default/output',
    'embedding_name': 'default_embedding',
    'clustering_method': 'kmeans',
    'num_clusters': 5,
    'other_param': 'unchanged'
}

# Create an argparse.Namespace with some overrides and some None values
args = argparse.Namespace(
    ids_path='/override/ids/path',
    embedding_path=None,
    output_dir='/override/output',
    embedding_name=None,
    clustering_method='hierarchical',
    num_clusters=10
)

ARGS = {
    'config': base_config,
    'args': args
}
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 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
apply_overrides Ran
ran; our contract draft was wrong, not the code
Richard-zrx/ADS/src/compute_difficulty.py
code served (permissive licence) · get_code("935fcc3e4436189a")
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 argparse
import torch

torch.manual_seed(0)

# Create a base config dictionary
config = {
    'model_name_or_path': 'bert-base-uncased',
    'input_path': '/tmp/input.txt',
    'output_dir': '/tmp/output',
    'batch_size': 16,
    'max_length': 512,
    'dtype': 'float32',
    'learning_rate': 1e-5,
    'num_epochs': 3
}

# Create an argparse.Namespace with override values
args = argparse.Namespace(
    model_name_or_path='gpt2',
    input_path='/tmp/input_override.txt',
    output_dir='/tmp/output_override',
    batch_size=32,
    max_length=256,
    dtype='float16',
    max_samples=1000
)

ARGS = {
    'config': config,
    'args': args
}
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 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
apply_overrides Ran
ran; our contract draft was wrong, not the code
Richard-zrx/ADS/src/export_cluster_examples.py
code served (permissive licence) · get_code("78858b5a31e7c77f")
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 argparse
from typing import Any

# Create a base config dictionary
config = {
    'assignments_path': None,
    'output_path': None,
    'top_k_nearest': 5,
    'top_k_random': 3,
    'other_setting': 'some_value'
}

# Create an argparse.Namespace with various override scenarios
args = argparse.Namespace(
    assignments_path='/path/to/assignments.txt',
    output_path='/path/to/output.txt',
    top_k_nearest=10,
    top_k_random=7
)

ARGS = {
    'config': config,
    'args': args
}
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 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
apply_overrides Ran
ran; our contract draft was wrong, not the code
Richard-zrx/ADS/src/extract_embeddings.py
code served (permissive licence) · get_code("b6fa95adbe2428b5")
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 argparse
import torch

torch.manual_seed(0)

# Create a minimal argparse.Namespace with all the expected attributes
args = argparse.Namespace(
    model_name_or_path="bert-base-uncased",
    input_path="/path/to/input.txt",
    output_dir="/path/to/output",
    batch_size=32,
    max_length=128,
    dtype="float32",
    pooling_methods="mean,max,cls",
    max_samples=1000
)

# Create a minimal config dict
config = {
    "model_name_or_path": "distilbert-base-uncased",
    "input_path": "/default/input.txt",
    "output_dir": "/default/output",
    "batch_size": 16,
    "max_length": 256,
    "dtype": "float16",
    "pooling_methods": ["cls"],
    "max_samples": 5000
}

ARGS = {
    "config": config,
    "args": args
}
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 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
apply_overrides Ran
ran; our contract draft was wrong, not the code
Richard-zrx/ADS/src/inspect_embeddings.py
code served (permissive licence) · get_code("47c1dc9fe9a63cf6")
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 argparse
from typing import Any

torch_imported = False
try:
    import torch
    torch_imported = True
except ImportError:
    pass

# Create a sample config dict
config = {
    'ids_path': '/default/ids/path.txt',
    'embedding_path': '/default/embedding/path.pt',
    'embedding_name': 'default_embedding',
    'output_dir': '/default/output',
    'other_key': 'other_value',
    'learning_rate': 0.001
}

# Create argparse.Namespace with some overrides
args = argparse.Namespace(
    ids_path='/override/ids/path.txt',
    embedding_path=None,  # This one is None, so won't override
    embedding_name='custom_embedding',
    output_dir='/override/output'
)

ARGS = {
    'config': config,
    'args': args
}
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 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
build_input_text Ran
ran; our contract draft was wrong, not the code
Richard-zrx/ADS/src/data_utils.py
code served (permissive licence) · get_code("77a7eca8a74d3263")
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 str. 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 json
import numpy as np
from typing import Any

# Construct a minimal example dict that satisfies the function's requirements
example = {
    'prompt': 'What is the capital of France?',
    'reward_model': {
        'ground_truth': 'Paris is the capital of France.'
    },
    'target': [
        {
            'content': 'Let me think step by step. France is a country in Europe. The capital of France is Paris.'
        }
    ]
}

ARGS = {
    'example': example,
    'text_field': 'prompt',
    'prompt_text_mode': 'all',
    'input_text_mode': 'question_only'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
ensure_dir Ran
ran; our contract draft was wrong, not the code
Richard-zrx/ADS/src/io_utils.py
code served (permissive licence) · get_code("25ba597a53fa8704")
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 PosixPath. 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
from pathlib import Path

# Create a temporary directory to use as a base
temp_base = tempfile.gettempdir()

# Construct valid arguments for ensure_dir
ARGS = {
    "path": str(Path(temp_base) / "test_ensure_dir_subdir")
}
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 PosixPath — a PosixPath output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a PosixPath 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
ensure_sample_id Ran
ran, honoured the contract we drafted
Richard-zrx/ADS/src/data_utils.py
code served (permissive licence) · get_code("eb3c5a08b9930b09")
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 numpy as np

# Test case 1: example with direct id_field
example_1 = {
    'id': 42,
    'text': 'sample'
}

# Test case 2: example with extra_info containing index
example_2 = {
    'name': 'test',
    'extra_info': {
        'index': 100
    }
}

# Test case 3: example with string id in extra_info
example_3 = {
    'data': 'value',
    'extra_info': {
        'index': 'sample_001'
    }
}

# Test case 4: example with no id info
example_4 = {
    'content': 'test'
}

# Test case 5: example with None values and fallback
example_5 = {
    'id': None,
    'extra_info': None
}

# Test case 6: example with float that is integer
example_6 = {
    'id': 7.0
}

# Test case 7: example with numpy integer
example_7 = {
    'id': np.int32(55)
}

# Select one example and parameters for the function call
ARGS = {
    'example': example_1,
    'idx': 0,
    'id_field': 'id'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — 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
load_config Ran
ran; our contract draft was wrong, not the code
Richard-zrx/ADS/src/cluster_embeddings.py
code served (permissive licence) · get_code("dec92090b9073869")
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 tempfile
import os
import yaml

# Create a temporary YAML config file
temp_dir = tempfile.mkdtemp()
config_path = os.path.join(temp_dir, "test_config.yaml")

# Write a valid YAML config file
config_data = {
    "model": "test_model",
    "batch_size": 32,
    "learning_rate": 0.001,
    "epochs": 10,
    "nested": {
        "key1": "value1",
        "key2": 42
    }
}

with open(config_path, 'w', encoding='utf-8') as f:
    yaml.dump(config_data, f)

ARGS = {
    "path": config_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
yaml — 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_config Ran
ran; our contract draft was wrong, not the code
Richard-zrx/ADS/src/compute_difficulty.py
code served (permissive licence) · get_code("d9faa034c888d1cf")
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 tempfile
import os
import yaml

# Create a temporary YAML config file
temp_dir = tempfile.mkdtemp()
config_path = os.path.join(temp_dir, 'test_config.yaml')

# Write a valid YAML config to the temporary file
test_config = {
    'model': 'test_model',
    'learning_rate': 0.001,
    'batch_size': 32,
    'layers': [64, 128, 256],
    'debug': True
}

with open(config_path, 'w', encoding='utf-8') as f:
    yaml.dump(test_config, f)

# Define arguments for load_config
ARGS = {
    'path': config_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
yaml — 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_config Ran
ran; our contract draft was wrong, not the code
Richard-zrx/ADS/src/extract_embeddings.py
code served (permissive licence) · get_code("74175bdc2e7f1d77")
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 tempfile
import os
import yaml

# Create a temporary YAML config file
temp_dir = tempfile.mkdtemp()
config_path = os.path.join(temp_dir, "test_config.yaml")

# Write a valid YAML config to the temporary file
test_config = {
    "model": "test_model",
    "learning_rate": 0.001,
    "batch_size": 32,
    "epochs": 10,
    "layers": [64, 128, 256]
}

with open(config_path, 'w', encoding='utf-8') as f:
    yaml.dump(test_config, f)

ARGS = {
    "config_path": config_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
yaml — 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 Ran
ran; our contract draft was wrong, not the code
Richard-zrx/ADS/src/model_utils.py
code served (permissive licence) · get_code("3ec62009af07711a")
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 Mock. 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 unittest.mock import Mock, MagicMock

torch.manual_seed(0)

# Create a mock model that behaves like AutoModel.from_pretrained would return
mock_model = Mock()
mock_model.eval = Mock(return_value=None)
mock_model.to = Mock(return_value=mock_model)

# Mock AutoModel.from_pretrained to return our mock model
import transformers
original_from_pretrained = transformers.AutoModel.from_pretrained
transformers.AutoModel.from_pretrained = Mock(return_value=mock_model)

ARGS = {
    "model_name_or_path": "gpt2",
    "torch_dtype": torch.float32,
    "device": torch.device("cpu"),
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, transformers — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Mock — a Mock 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
load_tokenizer Ran
ran; our contract draft was wrong, not the code
Richard-zrx/ADS/src/model_utils.py
code served (permissive licence) · get_code("a079eb7fd6d1a19a")
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 MockTokenizer. 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
from unittest.mock import MagicMock, patch

# Since we cannot download models from the network, we need to mock the
# AutoTokenizer.from_pretrained call to return a valid tokenizer-like object

class MockTokenizer:
    def __init__(self):
        self.pad_token = None
        self.eos_token = "<|endoftext|>"
        self.eos_token_id = 50256
        self.pad_token_id = None
        self.padding_side = 'left'

# Patch AutoTokenizer.from_pretrained to return our mock tokenizer
original_from_pretrained = AutoTokenizer.from_pretrained

def mock_from_pretrained(model_name_or_path, trust_remote_code=False):
    return MockTokenizer()

AutoTokenizer.from_pretrained = mock_from_pretrained

ARGS = {
    "model_name_or_path": "gpt2"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, transformers — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type MockTokenizer — a MockTokenizer output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a MockTokenizer 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
normalize_pooling_methods Ran
ran; our contract draft was wrong, not the code
Richard-zrx/ADS/src/io_utils.py
code served (permissive licence) · get_code("4207999c2265f336")
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 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
import numpy as np

ARGS = {
    'methods': ['last_token', 'mean']
}
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 list, ndim 1, dtype family float — 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
resolve_torch_dtype Ran
ran; our contract draft was wrong, not the code
Richard-zrx/ADS/src/model_utils.py
code served (permissive licence) · get_code("eb36e2d3da86f1ca")
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 dtype. 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

torch.manual_seed(0)

# The function takes a single parameter: dtype_name (a string)
# It maps string names to torch dtypes
# Valid inputs are: 'float32', 'fp32', 'float16', 'fp16', 'bfloat16', 'bf16'
# (case-insensitive, with whitespace stripped)

ARGS = {
    'dtype_name': 'float32'
}
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 dtype — a dtype output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dtype 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
setup_logger Ran
ran; our contract draft was wrong, not the code
Richard-zrx/ADS/src/logging_utils.py
code served (permissive licence) · get_code("3bd7521c54817c51")
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 Logger. 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 logging

ARGS = {
    'name': 'test_logger',
    'rank': 0,
    'world_size': 1
}
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 Logger — a Logger output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Logger 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
shard_output_paths Ran
ran; our contract draft was wrong, not the code
Richard-zrx/ADS/src/io_utils.py
code served (permissive licence) · get_code("2e3185902a515d73")
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 tempfile
from pathlib import Path

# Create a temporary directory for output
temp_dir = tempfile.mkdtemp()

ARGS = {
    'output_dir': temp_dir,
    'rank': 0,
    'pooling_methods': ['last_token', 'mean']
}
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 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 17 samples lifted from this paper's linked repositories, 17 carry a function identity and 0 do not. Of the 17 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 0 later) and 16 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.

FunctionAlso present inEarliest-datedOthers named
ensure_dir 2 other papers
2 dated earlier
2605.11301 (2026-05)
LatentRouter: Can We Choose the Right Multimodal Model Before Seeing Its Answer?
2605.28222 (2026-05)
Analyzing Quality-Latency-Resource Trade-offs in a Technical Documentation RAG Assistant Using LoRA Adaptation

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

Large Language Models (LLMs) achieve remarkable reasoning capabilities through reinforcement learning (RL) post-training. However, existing RL post-training commonly relies on uniform data sampling, which ignores the semantic structure of the training data and the changing capability of the training policy. To address these limitations, we propose Adaptive Data Scheduling (ADS), a dual-level data scheduling framework for pacing RL post-training that replaces uniform sampling with an adaptive distribution over semantic clusters and policy-boundary sample selection. At the cluster level, ADS organizes samples according to semantic patterns and maintains an adaptive inter-cluster distribution to solidify current training progress. At the sample level, ADS performs intra-cluster scheduling to continuously sample policy-boundary samples, which provides informative relative advantages. Experimental results across three LLMs and seven reasoning benchmarks demonstrate that ADS improves average accuracy by 5.2% over Group Relative Policy Optimization (GRPO). Notably, ADS consistently improves RL methods with different objective designs, highlighting its potential as a general data scheduling strategy for LLM RL post-training. The source code is available at: https://github.com/Richard-zrx/ADS.

For agents

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

get_harvested_code_for_paper("2606.22305")
get_code_for_paper("2606.22305")
have("2606.22305")

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

curl https://syntology.ai/api/ran/2606.22305.json

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

[![Syntology run record](https://syntology.ai/api/ran/2606.22305.svg)](https://syntology.ai/paper/2606.22305)

Connect an agent — have() is free.