Papers › OctoThinker: Mid-training Incentivizes Reinforcement Learning Scaling

OctoThinker: Mid-training Incentivizes Reinforcement Learning Scaling

25 Jun 2025arXiv:2506.20512archive 2025-07-28

Zengzhi Wang, Fan Zhou, Xuefeng Li, PengFei Liu

Different base language model families, such as Llama and Qwen, exhibit divergent behaviors during post-training with reinforcement learning (RL), especially on reasoning-intensive tasks. What makes a base language model suitable for reinforcement learning? Gaining deeper insight into this question is essential for developing RL-scalable foundation models of the next generation. In this work, we investigate how mid-training strategies shape RL dynamics, focusing on two representative model families: Qwen and Llama. Our study reveals that (1) high-quality mathematical corpora, such as MegaMath-Web-Pro, significantly improve both base model and RL performance, while existing alternatives (e.g., FineMath-4plus) fail to do so; (2) further adding QA-style data, particularly long chain-of-thought (CoT) reasoning examples, enhances RL outcomes, and instruction data further unlocks this effect; (3) while long-CoT improves reasoning depth, it can also induce verbosity of model responses and unstability of RL training, underscoring the importance of data formatting; (4) scaling mid-training consistently leads to stronger downstream RL performance. Building on these insights, we introduce a two-stage mid-training strategy, Stable-then-Decay, in which base models are first trained on 200B tokens with a constant learning rate, followed by 20B tokens across three CoT-focused branches with learning rate decay. This yields OctoThinker, a family of models demonstrating strong RL compatibility and closing the performance gap with more RL-friendly model families, i.e., Qwen. We hope our work will help shape pre-training strategies for foundation models in the RL era. To support further research, we release our open-source models along with a curated math reasoning-intensive corpus of over 70 billion tokens (i.e., MegaMath-Web-Pro-Max).

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 17 code samples harvested from 1 repository linked to this paper: 16 ran (of which 0 constructed an object rather than computing a result; 6 with no instrument failure: 1 honoured, 0 violated, 5 with no contract checked; 10 where Syntology's instrument failed) · 1 unverified. Where Syntology's instrument failed: 9 where our contract draft was wrong; 1 where the contract check could not run; failures of Syntology's instrument, not of the code.

By repository: official repository: 17 samples from 1 repository, 16 ran (6 with no instrument failure; 10 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.

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.

gair-nlp/octothinker officialmentioned in papermentioned on GitHubpytorchApache-2.0Syntology: 17 samples, 16 ran (10 instrument-failed) · 1 unverified 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.

Code Syntology ran Syntology

17 samples harvested: 16 ran (of which 0 constructed an object rather than computing a result; 6 with no instrument failure: 1 honoured, 0 violated, 5 with no contract checked; 10 where Syntology's instrument failed) · 1 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
9ran · our draft was wrong
1ran; the contract check could not run
5ran
1unverified

Licence: 4 of the 17 samples are pointer only, meaning Syntology does not serve that copy's text. For 4 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 gair-nlp/octothinker. “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.

gair-nlp/octothinker official repository 17 samples, 16 ran (10 instrument-failed) · 1 unverified
aggregate_metrics in 13 papers eval/summarize_results.py ran · our draft was wrong Apache-2.0 (permissive) · 8b9e2baf73bb9a6f · report fixture: signature paths:list/1/str · returned dict · level recorded 2026-09-25
aggregate_predictions in 5 papers eval/summarize_results.py ran · our draft was wrong fingerprinted Apache-2.0 (permissive) · 33425041e7f88b7f · report fixture: signature paths:list/1/str · returned list · level recorded 2026-09-25
extract_boxed_answers in 9 papers eval/data_processing/answer_extraction.py ran · our draft was wrong fingerprinted Apache-2.0 (permissive) here; the stored copy is pointer only · 6390dea763ce122e · report fixture: signature text:str/0/str · returned list · level recorded 2026-09-24
extract_program in 7 papers eval/eval/eval_utils.py ran · our draft was wrong fingerprinted Apache-2.0 (permissive) here; the stored copy is pointer only · 1d961161c71cc153 · report fixture: signature result:str/0/str, last_only:bool/0/bool · returned str · level recorded 2026-09-25
extract_program_output in 6 papers eval/data_processing/answer_extraction.py ran · our draft was wrong fingerprinted Apache-2.0 (permissive) here; the stored copy is pointer only · 0772e5020f6b0849 · report fixture: signature pred_str:str/0/str · returned str · level recorded 2026-09-24
generate_completions in 38 papers eval/eval/utils.py ran Apache-2.0 (permissive) · 73f280086b2714ce · report fixture: signature model:callable/0/?, tokenizer:callable/0/?, prompts:list/1/str, batch_size:int/0/int, stop_id_sequences:list/1/list, end_of_generation_id_sequence:list/1/int, disable_tqdm:bool/0/bool, max_new_tokens:int/0/int, num_return_sequences:int/0/int · returned tuple · level recorded 2026-09-25
get_next_word_predictions in 18 papers eval/eval/utils.py ran Apache-2.0 (permissive) · 19dce514835bd816 · report fixture: signature model:callable/0/?, tokenizer:callable/0/?, prompts:list/1/str, candidate_token_ids:list/1/int, batch_size:int/0/int, return_token_predictions:bool/0/bool, disable_tqdm:bool/0/bool · returned tuple · level recorded 2026-09-25
markup_question eval/run_subset_parallel.py ran Apache-2.0 (permissive) · de0231790ab22287 · report fixture: signature args:NoneType/0/?, item:dict/0/?, language:str/0/str, src:str/0/str, task:str/0/str · returned dict · level recorded 2026-09-25
normalize_numeric eval/eval/ocwcourses_eval_utils.py ran · honoured contract fingerprinted Apache-2.0 (permissive) · 1c2220bf6666327d · report fixture: signature s:str/0/str · returned float · level recorded 2026-09-25
normalize_symbolic_equation eval/eval/ocwcourses_eval_utils.py ran · our draft was wrong fingerprinted Apache-2.0 (permissive) · 73354091e7f44255 · report fixture: signature s:str/0/str · returned str · level recorded 2026-09-25
numeric_equality eval/eval/ocwcourses_eval_utils.py ran Apache-2.0 (permissive) · 4951ea903a7ffa37 · report fixture: signature n1:float/0/float, n2:float/0/float, threshold:float/0/float · returned bool · level recorded 2026-09-25
parse_question in 5 papers eval/eval/eval_utils.py ran Apache-2.0 (permissive) · 49e328786ffb3f8f · report fixture: signature example:dict/0/?, data_name:str/0/str · returned str · level recorded 2026-09-25
read_csv_file in 6 papers eval/summarize_results.py ran · our draft was wrong fingerprinted Apache-2.0 (permissive) · 5ae15dd305f032e2 · report fixture: signature file_path:str/0/str · returned DataFrame · level recorded 2026-09-25
read_data in 132 papers eval/utils.py ran · our draft was wrong Apache-2.0 (permissive) · 3deb39de933b985a · report fixture: signature path:str/0/str · returned list · level recorded 2026-09-25
shuffle in 35 papers eval/utils.py ran; the contract check could not run (RAISES) fingerprinted Apache-2.0 (permissive) · 3358374eed4646e4 · report fixture: signature data:list/1/int, seed:int/0/int · returned list · level recorded 2026-09-25
strip_string in 18 papers eval/data_processing/answer_extraction.py ran · our draft was wrong fingerprinted Apache-2.0 (permissive) here; the stored copy is pointer only · 5abed389274c4c03 · report fixture: signature string:str/0/str · returned str · level recorded 2026-09-24
score_completions in 11 papers eval/eval/utils.py unverified Apache-2.0 (permissive) · 8c63844c66973f8d · report

Tasks

Language ModelingLanguage ModellingMathReinforcement LearningReinforcement Learning (RL)reinforcement-learning

Results from the paper archive 2025-07-28

No leaderboard rows for this paper in the archive.

Methods

BASELLaMA

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