SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2407.18248 · ACL · 2024

Self-Training with Direct Preference Optimization Improves Chain-of-Thought Reasoning

Wei Lu, Shichen Li, Tianduo Wang

arXiv · PDF · Open in the Atlas

Code that ran

2 of 2 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; 1 failure 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/2407.18248.json · badge: /api/ran/2407.18248.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
tianduowang/dpo-st marked official in Papers with Code
Papers with Code archive
2 ran · 0 no recorded run
FunctionStatusWhere it lives
compare_similarity Ran
ran, honoured the contract we drafted
tianduowang/dpo-st/utils/make_dpo_data.py
code served (permissive licence) · get_code("28818764c84c2ec4")
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

class SimpleTokenizer:
    def tokenize(self, string):
        return string.split()

string1 = "hello world test"
string2 = "hello there world"
tokenizer = SimpleTokenizer()

ARGS = {
    "string1": string1,
    "string2": string2,
    "t": tokenizer
}
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
process_dpo_train Ran
ran; our contract draft was wrong, not the code
tianduowang/dpo-st/utils/make_dpo_data.py
code served (permissive licence) · get_code("f7bb9d98c9b915e0")
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 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 os
import json
import tempfile
from pathlib import Path

# Create a minimal tokenizer stand-in
class SimpleTokenizer:
    def tokenize(self, text):
        # Simple whitespace tokenization
        return text.lower().split()

# Create a minimal process function
def process_question(q):
    return q.strip()

def process_answer(a):
    return a.strip()

# Create temporary directory structure and files
temp_dir = tempfile.mkdtemp()
os.makedirs(f"{temp_dir}/model_outputs/dir1/train", exist_ok=True)
os.makedirs(f"{temp_dir}/model_outputs/dir2/train", exist_ok=True)

# Create sample DPO data files
sample_data_1 = [
    {
        "question": "What is the capital of France?",
        "gold_ans": "Paris is the capital of France",
        "positives": ["Paris is the capital", "The capital is Paris"],
        "negatives": ["London is the capital", "Berlin is the capital"]
    },
    {
        "question": "What is 2+2?",
        "gold_ans": "The answer is four",
        "positives": ["2+2 equals 4", "The sum is 4"],
        "negatives": ["The answer is five", "2+2 equals 5"]
    }
]

sample_data_2 = [
    {
        "question": "What is the capital of France?",
        "gold_ans": "Paris is the capital of France",
        "positives": ["Paris France capital", "Capital of France"],
        "negatives": ["Rome is capital", "Madrid is capital"]
    },
    {
        "question": "What is 2+2?",
        "gold_ans": "The answer is four",
        "positives": ["Four is the answer", "Sum equals four"],
        "negatives": ["Answer is three", "Sum is three"]
    }
]

# Write data files
with open(f"{temp_dir}/model_outputs/dir1/train/train_dpo_data.jsonl", 'w') as f:
    for item in sample_data_1:
        f.write(str(item) + '\n')

with open(f"{temp_dir}/model_outputs/dir2/train/train_dpo_data.jsonl", 'w') as f:
    for item in sample_data_2:
        f.write(str(item) + '\n')

# Change to temp directory so relative paths work
original_cwd = os.getcwd()
os.chdir(temp_dir)

ARGS = {
    "data_dir_lst": ["dir1", "dir2"],
    "tokenizer": SimpleTokenizer(),
    "process_question_func": process_question,
    "process_answer_func": process_answer
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
tqdm — 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

Also present in other papers' code

Of 2 samples lifted from this paper's linked repositories, 2 carry a function identity and 0 do not. Of the 2 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 2 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

Teaching small-scale language models to perform math reasoning is a valuable yet challenging task. Besides obtaining labeled data from human experts, one of the most common ways to collect high-quality data is by sampling from a larger and more powerful language model. Although previous works have demonstrated the effectiveness of this method, such a knowledge distillation paradigm can be costly and unstable, especially considering that many large language models, such as GPT-4 (Ope-nAI, 2023), are closed-source, proprietary, and their behaviors are unpredictable. In this work, to avoid relying on outputs from large models, we demonstrate that the reasoning abilities of small-scale language models can be enhanced through self-training, which involves training models with their own outputs. We also show that the conventional self-training can be further augmented by an alignment algorithm called Direct Preference Optimization (DPO) (Rafailov et al., 2023). We empirically found that models trained with the DPO objective are capable of making better generations that largely benefit multi-turn self-training. The experimental results show our models outperform the existing models with comparable sizes on the GSM8K benchmark with minimal resource requirements. 1

For agents

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

get_harvested_code_for_paper("2407.18248")
get_code_for_paper("2407.18248")
have("2407.18248")

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

curl https://syntology.ai/api/ran/2407.18248.json

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

[![Syntology run record](https://syntology.ai/api/ran/2407.18248.svg)](https://syntology.ai/paper/2407.18248)

Connect an agent — have() is free.