SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2505.03335 · 2025

Absolute Zero: Reinforced Self-play Reasoning with Zero Data

arXiv · PDF · Open in the Atlas

Code that ran

6 of 12 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, 1 violated it and 2 have no contract verdict; 2 failures of our instrument; 6 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/2505.03335.json · badge: /api/ran/2505.03335.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
LeapLabTHU/Absolute-Zero-Reasoner canonical (classifier label)
classifier basis not recorded on the link
6 ran · 6 no recorded run
FunctionStatusWhere it lives
count_unknown_letters_in_expr Ran
ran, honoured the contract we drafted
LeapLabTHU/Absolute-Zero-Reasoner/absolute_zero_reasoner/rewards/math_utils.py
code served (permissive licence) · get_code("368d3db6347269c3")
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 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

def count_unknown_letters_in_expr(expr: str):
    expr = expr.replace('sqrt', '')
    expr = expr.replace('frac', '')
    letters_in_expr = set([x for x in expr if x.isalpha()])
    return len(letters_in_expr)

ARGS = {
    'expr': 'sqrt(x) + frac(a, b) + y'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
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
extract_code Ran
ran; our contract draft was wrong, not the code
LeapLabTHU/Absolute-Zero-Reasoner/absolute_zero_reasoner/rewards/code_reward.py
pointer only (licence: MIT); the stored copy get_code serves, from Hsiang-1/MiniOpt, is pointer only (licence: NONE) · get_code("dd14950a6b76ffbc")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from Hsiang-1/MiniOpt, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

extract_code Ran
ran on a synthesized input
LeapLabTHU/Absolute-Zero-Reasoner/absolute_zero_reasoner/rewards/custom_evaluate.py
code served (permissive licence) · get_code("72c1465a54867eee")
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 re

def extract_code(completion: str, language: str='python') -> str:
    pattern = re.compile(f'```{language}\\n(.*?)```', re.DOTALL)
    matches = pattern.findall(completion)
    extracted_answer = matches[-1] if len(matches) >= 1 else ''
    return extracted_answer

# Setup code
completion_with_code = """
Some explanation here.


def hello():
    return "world"


More text.


def final_answer():
    return 42

"""

ARGS = {
    'completion': completion_with_code,
    'language': 'python'
}
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 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
mathd_normalize_answer Ran
ran; our contract draft was wrong, not the code
LeapLabTHU/Absolute-Zero-Reasoner/absolute_zero_reasoner/rewards/math_utils.py
code served (permissive licence) · get_code("ec578739f88adc22")
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 re
from typing import Optional

# Test cases for mathd_normalize_answer
test_cases = [
    # Basic string
    "42",
    # String with text wrapper
    "\\text{hello world}",
    # String with LaTeX fractions
    "\\frac{1}{2}",
    # String with newlines and special chars
    "3.14\n",
    # String with tfrac/dfrac
    "\\tfrac{3}{4}",
    # String with sqrt
    "\\sqrt{4}",
    # String with spaces
    "1 / 2",
    # String with = sign
    "x = 5",
    # Empty-like strings
    ".5",
    # None
    None,
]

ARGS = {
    "answer": "\\text{\\frac{1}{2}}"
}
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 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
process_livecodebench_execution Ran
ran on a synthesized input
LeapLabTHU/Absolute-Zero-Reasoner/absolute_zero_reasoner/data_construction/process_code_reasoning_data.py
code served (permissive licence) · get_code("c077759d0f068d30")
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 re

# Construct a valid row dict for process_livecodebench_execution
# The function expects:
# - row['problem']: a string containing Python code with function definitions
# - row['input']: a string containing a function call

# Create a sample problem with a function definition
problem_code = """
def calculate_sum(a, b):
    return a + b

def helper_func(x):
    return x * 2
"""

# Create a sample input that calls one of the functions
input_string = "calculate_sum(5, 3)"

# Build the row dict
row = {
    'problem': problem_code,
    'input': input_string
}

ARGS = {
    'row': row
}
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
should_allow_eval Ran
ran, violated the contract we drafted
LeapLabTHU/Absolute-Zero-Reasoner/absolute_zero_reasoner/rewards/math_utils.py
code served (permissive licence) · get_code("95f02454eddf082a")
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 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 re

BAD_SUBSTRINGS = ['^{', '^(']

BAD_REGEXES = ['\\^[0-9]+\\^', '\\^[0-9][0-9]+']

def count_unknown_letters_in_expr(expr: str):
    expr = expr.replace('sqrt', '')
    expr = expr.replace('frac', '')
    letters_in_expr = set([x for x in expr if x.isalpha()])
    return len(letters_in_expr)

def should_allow_eval(expr: str):
    if count_unknown_letters_in_expr(expr) > 2:
        return False
    for bad_string in BAD_SUBSTRINGS:
        if bad_string in expr:
            return False
    for bad_regex in BAD_REGEXES:
        if re.search(bad_regex, expr) is not None:
            return False
    return True

# Test cases for the function
ARGS = {
    'expr': '2*x + 3*y'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 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
add_imports Not yet run
unverified (harvested, no recorded run)
LeapLabTHU/Absolute-Zero-Reasoner/absolute_zero_reasoner/data_construction/process_code_reasoning_data.py
code served (permissive licence) · get_code("7dadfcd516d43505")
How it ran

no invocation recorded.

choice_answer_clean Not yet run
unverified (harvested, no recorded run)
LeapLabTHU/Absolute-Zero-Reasoner/absolute_zero_reasoner/rewards/custom_evaluate.py
code served (permissive licence) · get_code("0742e5f056e04404")
How it ran

no invocation recorded.

get_code_problem_generator_prompt Not yet run
unverified (harvested, no recorded run)
LeapLabTHU/Absolute-Zero-Reasoner/absolute_zero_reasoner/data_construction/prompts.py
code served (permissive licence) · get_code("a08febbb4b8dc4c6")
How it ran

no invocation recorded.

get_code_problem_predictor_prompt Not yet run
unverified (harvested, no recorded run)
LeapLabTHU/Absolute-Zero-Reasoner/absolute_zero_reasoner/data_construction/prompts.py
code served (permissive licence) · get_code("1b213fd26ce83e1b")
How it ran

no invocation recorded.

parse_to_ast Not yet run
unverified (harvested, no recorded run)
LeapLabTHU/Absolute-Zero-Reasoner/absolute_zero_reasoner/rewards/code_reward.py
code served (permissive licence) · get_code("48586c02e33616fc")
How it ran

no invocation recorded.

run_huggingface_download Not yet run
unverified (harvested, no recorded run)
LeapLabTHU/Absolute-Zero-Reasoner/absolute_zero_reasoner/utils/remove_think_qwen3_tokenizer.py
code served (permissive licence) · get_code("d1604f4eadcbe65a")
How it ran

no invocation recorded.

Also present in other papers' code

Of 12 samples lifted from this paper's linked repositories, 12 carry a function identity and 0 do not. Of the 12 distinct functions with an identity, 5 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 3 the same month, 5 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.

FunctionAlso present inEarliest-datedOthers named
should_allow_eval 13 other papers
4 dated earlier, 2 the same month, 7 later
2305.20050 (2023-05)
Let's Verify Step by Step
2406.12809 (2024-06)
Can Large Language Models Always Solve Easy Problems if They Can Solve Harder Ones?
2502.06703 (2025-02)
Can 1B LLM Surpass 405B LLM? Rethinking Compute-Optimal Test-Time Scaling
and 10 more not named here
count_unknown_letters_in_expr 12 other papers
3 dated earlier, 2 the same month, 7 later
2305.20050 (2023-05)
Let's Verify Step by Step
2502.06703 (2025-02)
Can 1B LLM Surpass 405B LLM? Rethinking Compute-Optimal Test-Time Scaling
2503.10460 (2025-03)
Light-R1: Curriculum SFT, DPO and RL for Long COT from Scratch and Beyond
and 9 more not named here
mathd_normalize_answer 12 other papers
1 dated earlier, 2 the same month, 9 later
2503.10460 (2025-03)
Light-R1: Curriculum SFT, DPO and RL for Long COT from Scratch and Beyond
2505.13379 (2025-05)
Thinkless: LLM Learns When to Think
2505.13417 (2025-05)
AdaptThink: Reasoning Models Can Learn When to Think
and 9 more not named here
extract_code 1 other paper
1 later
2606.25832 (2026-06)
MiniOpt: Reasoning to Model and Solve General Optimization Problems with Limited Resources
run_huggingface_download 1 other paper
1 later
2605.11666 (2026-05)
Evolutionary Task Discovery: Advancing Reasoning Frontiers via Skill Composition and Complexity Scaling

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

Reinforcement learning with verifiable rewards (RLVR) has shown promise in enhancing the reasoning capabilities of large language models by learning directly from outcome-based rewards. Recent RLVR works that operate under the zero setting avoid supervision in labeling the reasoning process, but still depend on manually curated collections of questions and answers for training. The scarcity of high-quality, human-produced examples raises concerns about the long-term scalability of relying on human supervision, a challenge already evident in the domain of language model pretraining. Furthermore, in a hypothetical future where AI surpasses human intelligence, tasks provided by humans may offer limited learning potential for a superintelligent system. To address these concerns, we propose a new RLVR paradigm called Absolute Zero, in which a single model learns to propose tasks that maximize its own learning progress and improves reasoning by solving them, without relying on any external data. Under this paradigm, we introduce the Absolute Zero Reasoner (AZR), a system that self-evolves its training curriculum and reasoning ability by using a code executor to both validate proposed code reasoning tasks and verify answers, serving as an unified source of verifiable reward to guide open-ended yet grounded learning. Despite being trained entirely without external data, AZR achieves overall SOTA performance on coding and mathematical reasoning tasks, outperforming existing zero-setting models that rely on tens of thousands of in-domain human-curated examples. Furthermore, we demonstrate that AZR can be effectively applied across different model scales and is compatible with various model classes.

For agents

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

get_harvested_code_for_paper("2505.03335")
get_code_for_paper("2505.03335")
have("2505.03335")

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

curl https://syntology.ai/api/ran/2505.03335.json

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

[![Syntology run record](https://syntology.ai/api/ran/2505.03335.svg)](https://syntology.ai/paper/2505.03335)

Connect an agent — have() is free.