SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2511.04902 · NeurIPS · 2025

You Need Reasoning to Learn Reasoning: The Limitations of Label-Free RL in Weak Base Models

Hossein Hajimirsadeghi, Shuvendu Roy, Mengyao Zhai, Golnoosh Samei

arXiv · PDF · Open in the Atlas

Code that ran

9 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; 3 honoured the contract we drafted, 1 violated it and 0 have no contract verdict; 5 failures of our instrument; 3 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/2511.04902.json · badge: /api/ran/2511.04902.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
BorealisAI/CuMa canonical (classifier label)
classifier basis: census
9 ran · 3 no recorded run
FunctionStatusWhere it lives
format_reward Ran
ran, honoured the contract we drafted
BorealisAI/CuMa/src/open_r1/rewards.py
pointer only (licence: NONE) · get_code("ce353d566f3f2c9f")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, BorealisAI/CuMa 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.

get_callbacks Ran
ran; our contract draft was wrong, not the code
BorealisAI/CuMa/src/open_r1/utils/callbacks.py
pointer only (licence: NONE); the stored copy get_code serves, from tuoyou-hao/BPD-GRPO, is pointer only (licence: NONE) · get_code("9b92c44e1bde6319")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, BorealisAI/CuMa 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.

get_param_count_from_repo_id Ran
ran, honoured the contract we drafted
BorealisAI/CuMa/src/open_r1/utils/hub.py
pointer only (licence: NONE); the stored copy get_code serves, from tuoyou-hao/BPD-GRPO, is pointer only (licence: NONE) · get_code("dcbd918d943fcaab")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, BorealisAI/CuMa 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.

nanmin Ran
ran, violated the contract we drafted
BorealisAI/CuMa/src/open_r1/intuitor_trainer.py
pointer only (licence: NONE); the stored copy get_code serves, from zxiangx/lc-r1, is pointer only (licence: NONE) · get_code("1fa45c1d5b9a7e0a")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, BorealisAI/CuMa 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.

nanstd Ran
ran; our contract draft was wrong, not the code
BorealisAI/CuMa/src/open_r1/intuitor_trainer.py
pointer only (licence: NONE) · get_code("a6cece2ffebe0f9c")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, BorealisAI/CuMa 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.

noise_reward Ran
ran, honoured the contract we drafted
BorealisAI/CuMa/src/open_r1/rewards.py
pointer only (licence: NOASSERTION) · get_code("bebd9517b1cc9b17")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, BorealisAI/CuMa (NOASSERTION). 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.

parse_questions Ran
ran; our contract draft was wrong, not the code
BorealisAI/CuMa/data_curation.py
pointer only (licence: NOASSERTION) · get_code("d615413e1a73313b")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, BorealisAI/CuMa (NOASSERTION). 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.

sanitize_example Ran
ran; our contract draft was wrong, not the code
BorealisAI/CuMa/data_curation.py
pointer only (licence: NOASSERTION) · get_code("b47fed0756359a18")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, BorealisAI/CuMa (NOASSERTION). 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.

split_tensor_dict Ran
ran; our contract draft was wrong, not the code
BorealisAI/CuMa/src/open_r1/intuitor_trainer.py
pointer only (licence: NONE) · get_code("efd67e2adf6b1ba5")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, BorealisAI/CuMa 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.

get_gpu_count_for_vllm Not yet run
unverified (harvested, no recorded run)
BorealisAI/CuMa/src/open_r1/utils/hub.py
pointer only (licence: NONE); the stored copy get_code serves, from tuoyou-hao/BPD-GRPO, is pointer only (licence: NONE) · get_code("4b0eb1a9216422d3")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, BorealisAI/CuMa 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. No invocation recorded.

get_provider Not yet run
unverified (harvested, no recorded run)
BorealisAI/CuMa/src/open_r1/utils/code_providers.py
pointer only (licence: NONE) · get_code("171250f85a491f40")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, BorealisAI/CuMa 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. No invocation recorded.

sample_examples Not yet run
unverified (harvested, no recorded run)
BorealisAI/CuMa/data_curation.py
pointer only (licence: NOASSERTION) · get_code("72792270fdf80bee")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, BorealisAI/CuMa (NOASSERTION). 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. 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, 8 are also present in other papers whose code we hold (8 with at least one paper dated earlier, 0 the same month, 8 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.

FunctionAlso present inEarliest-datedOthers named
nanstd 7 other papers
3 dated earlier, 4 later
2402.16302 (2024-02)
Graph Diffusion Policy Optimization
2505.17534 (2025-05)
Co-Reinforcement Learning for Unified Multimodal Understanding and Generation
2506.14755 (2025-06)
Optimizing Length Compression in Large Reasoning Models
and 4 more not named here
split_tensor_dict 7 other papers
3 dated earlier, 4 later
2402.16302 (2024-02)
Graph Diffusion Policy Optimization
2505.17534 (2025-05)
Co-Reinforcement Learning for Unified Multimodal Understanding and Generation
2506.14755 (2025-06)
Optimizing Length Compression in Large Reasoning Models
and 4 more not named here
format_reward 5 other papers
2 dated earlier, 3 later
2506.20639 (2025-06)
DiffuCoder: Understanding and Improving Masked Diffusion Models for Code Generation
2507.02834 (2025-07)
ExPO: Unlocking Hard Reasoning with Self-Explanation-Guided Reinforcement Learning
2602.02258 (2026-02)
Alignment-Aware Model Adaptation via Feedback-Guided Optimization
and 2 more not named here
get_callbacks 5 other papers
2 dated earlier, 3 later
2503.16219 (2025-03)
Reinforcement Learning for Reasoning in Small LLMs: What Works and What Doesn't
2507.02834 (2025-07)
ExPO: Unlocking Hard Reasoning with Self-Explanation-Guided Reinforcement Learning
2601.19404 (2026-01)
RPO:Reinforcement Fine-Tuning with Partial Reasoning Optimization
and 2 more not named here
nanmin 5 other papers
1 dated earlier, 3 later, 1 undated
2506.14755 (2025-06)
Optimizing Length Compression in Large Reasoning Models
2601.04700 (2026-01)
PRISM: A Unified Framework for Post-Training LLMs Without Verifiable Rewards
2605.15726 (2026-05)
Nudging Beyond the Comfort Zone: Efficient Strategy-Guided Exploration for RLVR
and 2 more not named here
get_gpu_count_for_vllm 3 other papers
2 dated earlier, 1 later
2503.16219 (2025-03)
Reinforcement Learning for Reasoning in Small LLMs: What Works and What Doesn't
2507.02834 (2025-07)
ExPO: Unlocking Hard Reasoning with Self-Explanation-Guided Reinforcement Learning
2606.21498 (2026-06)
Balancing Performance and Diversity in GRPO Autoregressive Text-to-Image Post-Training
get_param_count_from_repo_id 3 other papers
2 dated earlier, 1 later
2503.16219 (2025-03)
Reinforcement Learning for Reasoning in Small LLMs: What Works and What Doesn't
2507.02834 (2025-07)
ExPO: Unlocking Hard Reasoning with Self-Explanation-Guided Reinforcement Learning
2606.21498 (2026-06)
Balancing Performance and Diversity in GRPO Autoregressive Text-to-Image Post-Training
get_provider 3 other papers
2 dated earlier, 1 later
2506.20639 (2025-06)
DiffuCoder: Understanding and Improving Masked Diffusion Models for Code Generation
2507.02834 (2025-07)
ExPO: Unlocking Hard Reasoning with Self-Explanation-Guided Reinforcement Learning
2606.12900 (2026-06)
Zero-source LLM Hallucination Detection with Human-like Criteria Probing

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

Recent advances in large language models have demonstrated the promise of unsupervised reinforcement learning (RL) methods for enhancing reasoning capabilities without external supervision. However, the generalizability of these label-free RL approaches to smaller base models with limited reasoning capabilities remains unexplored. In this work, we systematically investigate the performance of label-free RL methods across different model sizes and reasoning strengths, from 0.5B to 7B parameters. Our empirical analysis reveals critical limitations: label-free RL is highly dependent on the base model's pre-existing reasoning capability, with performance often degrading below baseline levels for weaker models. We find that smaller models fail to generate sufficiently long or diverse chain-of-thought reasoning to enable effective self-reflection, and that training data difficulty plays a crucial role in determining success. To address these challenges, we propose a simple yet effective method for label-free RL that utilizes curriculum learning to progressively introduce harder problems during training and mask no-majority rollouts during training. Additionally, we introduce a data curation pipeline to generate samples with predefined difficulty. Our approach demonstrates consistent improvements across all model sizes and reasoning capabilities, providing a path toward more robust unsupervised RL that can bootstrap reasoning abilities in resource-constrained models. We make our code available at https://github.com/BorealisAI/CuMa.

For agents

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

get_harvested_code_for_paper("2511.04902")
get_code_for_paper("2511.04902")
have("2511.04902")

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

curl https://syntology.ai/api/ran/2511.04902.json

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

[![Syntology run record](https://syntology.ai/api/ran/2511.04902.svg)](https://syntology.ai/paper/2511.04902)

Connect an agent — have() is free.