Papers › On the Emergence of Thinking in LLMs I: Searching for the Right Intuition

On the Emergence of Thinking in LLMs I: Searching for the Right Intuition

10 Feb 2025arXiv:2502.06773archive 2025-07-28

Guanghao Ye, Khiem Duc Pham, Xinzhi Zhang, Sivakanth Gopi, Baolin Peng, Beibin Li, Janardhan Kulkarni, Huseyin A. Inan

Recent AI advancements, such as OpenAI's new models, are transforming LLMs into LRMs (Large Reasoning Models) that perform reasoning during inference, taking extra time and compute for higher-quality outputs. We aim to uncover the algorithmic framework for training LRMs. Methods like self-consistency, PRM, and AlphaZero suggest reasoning as guided search. We ask: what is the simplest, most scalable way to enable search in LLMs? We propose a post-training framework called Reinforcement Learning via Self-Play (RLSP). RLSP involves three steps: (1) supervised fine-tuning with human or synthetic demonstrations of the reasoning process, (2) using an exploration reward signal to encourage diverse and efficient reasoning behaviors, and (3) RL training with an outcome verifier to ensure correctness while preventing reward hacking. Our key innovation is to decouple exploration and correctness signals during PPO training, carefully balancing them to improve performance and efficiency. Empirical studies in the math domain show that RLSP improves reasoning. On the Llama-3.1-8B-Instruct model, RLSP can boost performance by 23% in MATH-500 test set; On AIME 2024 math problems, Qwen2.5-32B-Instruct improved by 10% due to RLSP. However, a more important finding of this work is that the models trained using RLSP, even with the simplest exploration reward that encourages the model to take more intermediate steps, showed several emergent behaviors such as backtracking, exploration of ideas, and verification. These findings demonstrate that RLSP framework might be enough to enable emergence of complex reasoning abilities in LLMs when scaled. Lastly, we propose a theory as to why RLSP search strategy is more suitable for LLMs inspired by a remarkable result that says CoT provably increases computational power of LLMs, which grows as the number of steps in CoT.

PaperPDFCodeCode Syntology ran

In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

Syntology 13 code samples harvested from 1 repository linked to this paper: 11 ran (of which 0 constructed an object rather than computing a result; 6 with no instrument failure: 1 honoured, 2 violated, 3 with no contract checked; 5 where Syntology's instrument failed) · 2 unverified. Where Syntology's instrument failed: 5 where our contract draft was wrong; failures of Syntology's instrument, not of the code.

By repository: official repository: 13 samples from 1 repository, 11 ran (6 with no instrument failure; 5 where Syntology's instrument failed). The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

No sample is linked to this paper from 2 of its repositories that Syntology harvested for other papers:

Levels written between 2026-09-24 and 2026-09-25 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

GuanghaoYe/Emergence-of-Thinking officialmentioned on GitHubpytorchApache-2.0Syntology: 13 samples, 11 ran (5 instrument-failed) · 2 unverified report
OpenLLMAI/OpenLLaMA2 mentioned on GitHubpytorchApache-2.0Syntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Logic-RL: Unleashing LLM Reasoning with Rule-Based Reinforcement Learning (arXiv:2502.14768; that paper's own run record: 15 ran (of which 0 constructed an object rather than computing a result; 10 with no instrument failure: 1 honoured, 1 violated, 8 with no contract checked; 5 where Syntology's instrument failed) · 8 unverified); nothing checks that they implement this paper's method. report
OpenLLMAI/OpenRLHF mentioned on GitHubpytorchApache-2.0Syntology: not harvested report
openrlhf/openrlhf mentioned on GitHubpytorchApache-2.0Syntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to A Regret Minimization Framework on Preference Learning in Large Language Models (arXiv:2606.09124; that paper's own run record: 15 ran (of which 0 constructed an object rather than computing a result; 7 with no instrument failure: 1 honoured, 1 violated, 5 with no contract checked; 8 where Syntology's instrument failed) · 3 unverified); nothing checks that they implement this paper's method. report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

A paper named beside a repository with no sample linked to this paper is shown with that paper's own run record, not this paper's: “ran” means executed on a synthesized input, not that the code is correct, and “where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Papers are listed in arXiv-id order, at most three per repository.

Code Syntology ran Syntology

13 samples harvested: 11 ran (of which 0 constructed an object rather than computing a result; 6 with no instrument failure: 1 honoured, 2 violated, 3 with no contract checked; 5 where Syntology's instrument failed) · 2 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page last changed ; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

1ran · honoured contract
2ran · violated contract
5ran · our draft was wrong
3ran
2unverified

Licence: 9 of the 13 samples are pointer only, meaning Syntology does not serve that copy's text. For 9 of them, this paper's copy is under a permissive licence and cleared, but the stored copy of the same code that Syntology serves is not; the licence label's tooltip gives Syntology's recorded reason. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from GuanghaoYe/Emergence-of-Thinking. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.

Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of six reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity.

GuanghaoYe/Emergence-of-Thinking official repository 13 samples, 11 ran (5 instrument-failed) · 2 unverified
choice_answer_clean in 17 papers evaluation/grader.py ran · our draft was wrong fingerprinted Apache-2.0 (permissive) here; the stored copy is pointer only · 809c0178a7ae90ff · report fixture: signature pred:str/0/str · returned str · level recorded 2026-09-25
extract_multi_choice_answer in 13 papers evaluation/parser.py ran · our draft was wrong fingerprinted Apache-2.0 (permissive) here; the stored copy is pointer only · ed2f4689128f9885 · report fixture: signature pred_str:str/0/str · returned str · level recorded 2026-09-25
is_choice evaluation/eval_math_data_parallel.py ran Apache-2.0 (permissive) · 32a0caa6856c96c1 · report fixture: signature answer:str/0/str · returned bool · level recorded 2026-09-25
is_digit in 33 papers evaluation/grader.py ran · violated contract Apache-2.0 (permissive) here; the stored copy is pointer only · fb37d3c84b18e368 · report fixture: signature num:str/0/str · returned bool · level recorded 2026-09-25
is_execution_success evaluation/trajectory.py ran · violated contract Apache-2.0 (permissive) here; the stored copy is pointer only · 581f3949cfdcffa2 · report fixture: signature output:str/0/str · returned bool · level recorded 2026-09-25
is_multi_choice in 9 papers evaluation/eval_math_data_parallel.py ran Apache-2.0 (permissive) · 17fa07643a5a08e5 · report fixture: signature answer:str/0/str · returned bool · level recorded 2026-09-25
lower_keys in 7 papers evaluation/utils.py ran · our draft was wrong Apache-2.0 (permissive) here; the stored copy is pointer only · 0af0b06d6cf22524 · report fixture: signature example:dict/0/? · returned dict · level recorded 2026-09-25
majority_voting evaluation/evaluate.py ran Apache-2.0 (permissive) · d12f00f6af3908b6 · report fixture: signature samples:list/1/dict · returned float · level recorded 2026-09-24
parse_digits in 20 papers evaluation/grader.py ran · honoured contract fingerprinted Apache-2.0 (permissive) here; the stored copy is pointer only · 7d4b7f8c3875b126 · report fixture: signature num:str/0/str · returned float · level recorded 2026-09-25
text_to_trajectory evaluation/trajectory.py ran · our draft was wrong fingerprinted Apache-2.0 (permissive) here; the stored copy is pointer only · 7e48538d19cc0d8f · report fixture: signature traj_str:str/0/str · returned list · level recorded 2026-09-25
trajectory_to_text evaluation/trajectory.py ran · our draft was wrong Apache-2.0 (permissive) here; the stored copy is pointer only · 0d7ae684ec6c7537 · report fixture: signature trajectory:list/1/dict · returned str · level recorded 2026-09-25
generate_completions in 38 papers evaluation/model_utils.py unverified Apache-2.0 (permissive) · 25f1381fe69df071 · report
load_hf_lm_and_tokenizer in 21 papers evaluation/model_utils.py unverified Apache-2.0 (permissive) here; the stored copy is pointer only · ee71a94d9dd1b8e9 · report

Tasks

Math

Results from the paper archive 2025-07-28

No leaderboard rows for this paper in the archive.

Methods

AlphaZeroEntropy RegularizationPPO

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections