Pith. sign in

REVIEW 4 major objections 6 minor 3 cited by

Scaling Up RL: Unlocking Diverse Reasoning in LLMs via Prolonged Training

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper claims that prolonged reinforcement learning, stabilized by a small KL penalty and periodic reference-policy resets, lifts a 1.5B model's reasoning across math, code, puzzles, STEM, and instruction-following to near-specialist…

desk verdict The reference-policy reset is a genuinely useful idea, but the validation/evaluation overlap means the headline gains are not yet established as generalization. read the letter →

arxiv 2507.12507 v1 pith:26I3BZ5X submitted 2025-07-16 cs.LG cs.AIcs.CL

Mingjie Liu , Shizhe Diao , Jian Hu , Ximing Lu , Xin Dong , Hao Zhang , Alexander Bukharin , Shaokun Zhang
show 14 more authors
Jiaqi Zeng Makesh Narsimhan Sreedhar Gerald Shen David Mosallanezhad Di Zhang Jonas Yang June Yang Oleksii Kuchaiev Guilin Liu Zhiding Yu Pavlo Molchanov Yejin Choi Jan Kautz Yi Dong
This is my paper
classification cs.LGcs.AIcs.CL
keywords reinforcementlearningGRPOverifiablerewardsKLregularizationreferencepolicyresetentropycollapsereasoningmodelssmalllanguage
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that prolonged reinforcement learning can keep improving a small language model across many reasoning domains, rather than plateauing early, if the training recipe includes a small KL-divergence penalty and periodic resets of the reference policy. Starting from a 1.5B reasoning checkpoint, the authors train on verifiable-reward tasks in math, code, STEM, logic puzzles, and instruction following, and report average gains over the base model of 14.7% on math, 13.9% on coding, 54.8% on logic puzzles, 25.1% on STEM, and 18.1% on instruction following. The result matters because it suggests that scaling training time on a modest model, with careful stabilization, can bridge the gap to domain-specialized systems.

What carries the argument

The load-bearing mechanism is the reference-policy reset: every so often the KL anchor $\pi_{\text{ref}}$ is hard-reset to a recent snapshot of the online policy, and optimizer states are reinitialized. This prevents the KL penalty, whose coefficient is small ($\beta = 10^{-4}$), from eventually dominating the loss and freezing learning, while still preserving the entropy-stabilizing effect of KL regularization against a fixed anchor. The other machinery is a GRPO objective with decoupled clipping bounds ($\epsilon_{low}=0.2$, $\epsilon_{high}=0.4$) and dynamic sampling that filters prompts where all rollouts succeed or all fail, which raises reward-signal density.

What would settle it

Train the same 1.5B model with the same budget but monitor and schedule resets using a validation set disjoint from all final evaluation benchmarks; if the reported gains on math, code, and logic puzzles disappear or shrink to noise, the central claim that these techniques unlock long-term improvement is not supported.

Watch

Extended reading notes

Core claim

The paper's central claim is that the limiting factor in RL for reasoning models is not model size but training stability over long horizons. With verifiable rewards, decoupled clipping, dynamic sampling, a small controlled KL penalty ($\beta = 10^{-4}$), and periodic hard resets of the reference policy and optimizer states, a 1.5B model continues to improve across diverse tasks for roughly sixteen thousand GPU-hours. Compared with its starting checkpoint, the model gains 14.7% on math, 13.9% on coding, 54.8% on logic puzzles, 25.1% on STEM, and 18.1% on instruction-following, and remains competitive with two domain-specialized 1.5B models trained for single domains.

Load-bearing premise

The training recipe's resets and hyperparameter choices were monitored against a validation blend that includes slices of the final evaluation benchmarks, so the reported gains may depend on peeking at the test distribution.

Editorial extensions

If this is right

  • Small-scale models can serve as test beds for RL recipes that later transfer to larger models, since stability techniques rather than scale drive the gains.
  • Removing KL regularization entirely, as some recent recipes do, may be wrong when starting from a strong chain-of-thought checkpoint; a small penalty can be needed.
  • Reference-policy resets give a practical intervention when validation plateaus or KL spikes, extending useful training without changing data.
  • Diverse verifiable-reward data across domains produces a generalist that matches single-domain specialists at the same size.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A fully blind reset schedule is the obvious next test: if the same gains appear without validation-guided intervention, the recipe is robust; if not, the evaluation blend is doing much of the work.
  • The staged training runs with human monitoring resemble an adaptive curriculum, so automating the detection of KL spikes or entropy collapse could make the method more reproducible.
  • Reference-policy reset is analogous to warm restarts in optimization and may reduce the need for very high rollout temperatures late in training.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. This paper reports on prolonged reinforcement learning of a 1.5B language model (DeepSeek-R1-Distill-Qwen-1.5B) across math, code, STEM, logical puzzles, and instruction-following tasks. The authors propose controlled KL regularization, decoupled clipping and dynamic sampling from DAPO, and periodic resets of the reference policy and optimizer states. They report average gains over the base model of +14.7% on math, +13.9% on coding, +54.8% on logic puzzles, +25.1% on STEM, and +18.1% on instruction-following, and competitive performance with domain-specialized 1.5B models. The paper includes ablation studies on temperature, clipping, dynamic sampling, entropy mitigation strategies, and reset.

Significance. If the results hold, this is a useful engineering contribution: it demonstrates that a 1.5B model can be trained for a long horizon on diverse verifiable-reward tasks with careful tuning, and it provides detailed training dynamics (KL, entropy, response length). The open-sourced model and the stage-by-stage recipe are valuable for reproducibility. The main weakness is that the training recipe is selected using a validation blend that overlaps with the final evaluation benchmarks, so the reported quantitative improvements are not yet established as generalization results.

major comments (4)
  1. [Section 4.2, 4.3] The validation blend described in Section 4.2 includes subsets of AIME2024, Codeforces, GPQA-diamond, IFEval, and graph_color from Reasoning Gym, and Section 4.2 explicitly states that this validation signal is used to decide hard resets, rollout-count changes, context-window changes, and reward shaping. These same benchmarks (or subsets thereof) are used in the final evaluation in Section 4.3, so the reported gains on them are at least partly a selection artifact rather than evidence of generalization. The overlap is most direct for logic puzzles: Section 2.4 reports 9600 validation samples across 96 tasks (100 per task), and Section 4.3 reports reserving 100 samples per task as test sets, so the two sets are likely identical. Please report results on benchmarks that were not used for any training decision, and disclose explicitly which evaluation sets overlap with the validation blend.
  2. [Section 5.3] The claim that periodic reference-policy resets are a critical component is not backed by a controlled ablation. The evidence in Figure 6 is a single extended training run that degrades on Codeforces, followed by a reset that also involves reinitializing optimizer states; in Runs 2, 4-7, resets are accompanied by changes in rollout count, context length, and reward shaping. This confound means the reader cannot attribute the recovery to the reset itself. A matched-compute comparison with and without reset, holding all other factors fixed, is needed to support the 'critical component' wording.
  3. [Section 4.4] All quantitative results come from a single training trajectory, and no error bars, confidence intervals, or multiple seeds are reported. The pass@1 estimates are computed from 16 samples per prompt, which on small benchmarks (e.g., AIME2024 has 30 problems) yields substantial variance. Without per-benchmark numbers with uncertainty, the aggregate improvements in the abstract and conclusion cannot be assessed for statistical significance.
  4. [Section 5.2] The ablations of decoupled clipping and dynamic sampling in Figure 5 appear to be performed on a small-scale AIME2024 validation setting, not on the final multi-stage recipe. The paper should state whether these conclusions transfer to the full recipe, and should provide error bars or multiple seeds for the reported 'clear improvements', since Figure 5(c) in particular shows only a single trajectory.
minor comments (6)
  1. [Abstract/Conclusion vs Section 4.4] The reported math improvement is inconsistent: the abstract and conclusion state +14.7%, while Section 4.4 says 'average improvement of 15.7%'. Please reconcile.
  2. [Section 4.3] Reference [31] is cited for IFEval, but [31] is 'Online difficulty filtering for reasoning oriented reinforcement learning' and is not the IFEval paper; IFEval is [9]. Please correct the citation.
  3. [Section 4.1] There are several typos: 'levearge' should be 'leverage', 'penaly' should be 'penalty', and '4 8 x NVIDIA-H100-80GB nodes' should likely read '48 x NVIDIA-H100-80GB nodes'. Section 4.4 has 'context windo' for 'context window', and Section 3 intro has 'instatility' for 'instability'.
  4. [Section 4.4] Please clarify how pass@1 is estimated from 16 samples; state whether an unbiased estimator is used or whether it is the mean per-prompt success rate.
  5. [Section 2.4] The text says 'approximately 100 tasks' while Section 4.3 says 96 tasks; consider stating '96 tasks' consistently in both places.
  6. [Section 5.1] In the phrase 'properly higher temperatures enable broader behavioral exploration', 'properly' is unclear; consider 'appropriately higher temperatures'.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the reported gains are empirical measurements rather than derivations; the main caveat is that Section 4.2's validation blend overlaps with the final evaluation benchmarks, which is a generalization risk but not a circular step.

full rationale

The paper does not derive predictions from fitted parameters or import a uniqueness theorem. Its claims are of the form 'after prolonged RL with these components, benchmark X improves by Y%', which are measured comparisons against DeepSeek-R1-Distill-Qwen-1.5B, DeepScaleR-1.5B, and DeepCoder-1.5B. The training techniques (KL penalty, decoupled clipping, dynamic sampling, reference resets) are validated through ablations reported in Section 5; none of the ablations equates its output to its input by construction. The only self-citation is the use of NVIDIA's Llama-Nemotron synthetic instruction-following data [8] and the IFEval-style setup; that citation is a data source, not a load-bearing theorem, so it does not make the central claim circular. The most serious caveat is in Section 4.2: the validation blend used to guide resets and hyperparameter changes explicitly includes subsets of the final evaluation benchmarks (AIME2024, Codeforces, GPQA-diamond, IFEval, graph_color), and Section 4.3 evaluates on those same benchmarks. For logic puzzles, Section 2.4 states 9600 validation samples across 96 tasks, while Section 4.3 reserves 100 samples per task as test sets (96 x 100 = 9600), so the text does not establish that the reported +54.8% logic-puzzle gain is on a disjoint set. This is a benchmark-selection/overfitting concern that should be addressed with a truly held-out evaluation, but it is not a circular derivation in the sense used here: no quantity is being predicted from itself, and the paper is transparent about the overlap. Score 1 reflects the minor self-citation and the acknowledged validation/evaluation overlap, not a reduction of the central claim to its inputs.

Assumptions & free parameters 8 free parameters · 6 assumptions · 0 invented entities

The paper introduces no new theoretical entities. Its central claim rests on a set of hand-tuned hyperparameters, a capable base model, and the assumption that verifiable rewards and benchmark evaluations are trustworthy. The reference-policy reset is a procedural mechanism rather than an invented entity.

free parameters (8)
  • KL penalty coefficient beta = 1e-4
    Chosen via ablations (Section 5.4); controls distance to reference policy.
  • Decoupled clipping thresholds = epsilon_low=0.2, epsilon_high=0.4
    Selected in Section 5.2 ablation to improve AIME2024 validation.
  • Rollout sampling temperature = 1.2
    Selected via early/late training ablations (Section 5.1).
  • Number of sampled responses per prompt = 16 (8k context) and 32 (Runs 6-7), then 16 (Run 8)
    Chosen for stability and throughput.
  • Learning rate = 2e-6
    Constant; listed in Section 4.1.
  • Dynamic sampling thresholds = exclude prompts with accuracy 0 or 1
    Adopted from DAPO to increase signal density.
  • Reference policy reset schedule = triggered by validation decline or KL spike
    Hand-tuned, not a fixed schedule; the criterion is based on validation metrics.
  • Reward shaping penalty for non-terminating responses = penalty amount not specified in the text
    Introduced in Runs 4 and 5 without specifying the penalty magnitude.
assumptions (6)
  • standard math GRPO advantage estimation uses group scores instead of a critic (Eq. 1-2).
    Adopted from DeepSeekMath and used as the foundation; not proved in this paper.
  • standard math The unbiased KL divergence estimator in Eq. 5 is correct.
    Cites Schulman's approximation; used without proof.
  • domain assumption Verifiable reward tasks provide objective, non-hackable training signal.
    Stated in Sections 1 and 2; the paper relies on this to attribute gains to reasoning rather than reward hacking.
  • domain assumption DeepSeek-R1-Distill-Qwen-1.5B is a capable CoT base model suitable for starting RL.
    Used as the initial policy; the authors note the KL finding may not transfer to non-CoT base models (Section 3.3.1).
  • domain assumption The filtered datasets (DeepScaleR math, Eurus code, SCP-116K STEM, Reasoning Gym puzzles, synthetic IFEval) are correctly verified and representative.
    Data quality is asserted but not independently audited; the SCP-116K filtering uses GPT-4o as judge.
  • domain assumption The evaluation benchmarks measure the claimed reasoning abilities.
    Used to report improvements; the validation blends overlap these benchmarks.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Scaling Up RL: Unlocking Diverse Reasoning in LLMs via Prolonged Training." pith.science (2026). https://pith.science/paper/26I3BZ5X

@misc{pith2026250712507,
  author       = {Pith},
  title        = {Pith review of: Scaling Up RL: Unlocking Diverse Reasoning in LLMs via Prolonged Training},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/26I3BZ5X}},
  note         = {Machine review of arXiv:2507.12507}
}
read the original abstract

Recent advancements in reasoning-focused language models such as OpenAI's O1 and DeepSeek-R1 have shown that scaling test-time computation-through chain-of-thought reasoning and iterative exploration-can yield substantial improvements on complex tasks like mathematics and code generation. These breakthroughs have been driven by large-scale reinforcement learning (RL), particularly when combined with verifiable reward signals that provide objective and grounded supervision. In this report, we investigate the effects of prolonged reinforcement learning on a small language model across a diverse set of reasoning domains. Our work identifies several key ingredients for effective training, including the use of verifiable reward tasks, enhancements to Group Relative Policy Optimization (GRPO), and practical techniques to improve training stability and generalization. We introduce controlled KL regularization, clipping ratio, and periodic reference policy resets as critical components for unlocking long-term performance gains. Our model achieves significant improvements over strong baselines, including +14.7% on math, +13.9% on coding, and +54.8% on logic puzzle tasks. To facilitate continued research, we release our model publicly.

Discussion (0). Sign in to comment.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Cognitive Pivot Points and Visual Anchoring: Unveiling and Rectifying Hallucinations in Multimodal Reasoning Models

    cs.AI 2026-04 unverdicted novelty 6.0 of 10

    Multimodal reasoning models hallucinate at high-entropy cognitive bifurcation points due to loss of visual semantic anchoring, and the V-STAR training paradigm with HVAR rewards and FRM reflection mitigates this by re...

  2. Learning to Reason Efficiently with Discounted Reinforcement Learning

    cs.LG 2025-10 conditional novelty 6.0 of 10

    Discounted correctness rewards in GRPO post-training shorten LLM reasoning traces by roughly 13-40% while matching pass@1 accuracy, supported by Blackwell-optimality analysis for restricted policy classes.

  3. Scaling DRL for Decision Making: A Survey on Data, Network, and Training Budget Strategies

    cs.LG 2025-08 conditional novelty 4.0 of 10

    A survey that categorizes deep reinforcement learning scaling strategies into data, network, and training budget dimensions and outlines challenges for scaling DRL systems.

Reference graph

Works this paper leans on

33 extracted references · 15 canonical work pages · cited by 3 Pith papers

  1. [1]

    Openai o1 system card.arXiv preprint arXiv:2412.16720 , 2024

    Aaron Jaech, Adam Kalai, Adam Lerer, Adam Richardson, Ahmed El-Kishky, Aiden Low, Alec Helyar, Aleksander Madry, Alex Beutel, Alex Carney, et al. Openai o1 system card.arXiv preprint arXiv:2412.16720 , 2024

    arXiv 2024

  2. [2]

    Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning.arXiv preprint arXiv:2501.12948, 2025

    Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, et al. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning.arXiv preprint arXiv:2501.12948, 2025

    arXiv 2025

  3. [3]

    Tang, Manan Roongta, Colin Cai, Jef- frey Luo, Li Erran Li, Raluca Ada Popa, and Ion Stoica

    Michael Luo, Sijun Tan, Justin Wong, Xiaoxiang Shi, William Y. Tang, Manan Roongta, Colin Cai, Jef- frey Luo, Li Erran Li, Raluca Ada Popa, and Ion Stoica. Deepscaler: Surpassing o1-preview with a 1.5b model by scaling rl.https://pretty-radio-b75 .notion.site/DeepScaleR-Surpassing-O1-Preview-with-a-1- 5B-Model-by-Scaling-RL-19681902c1468005bed8ca303013a4e...

    work page 2025

  4. [4]

    Dapo: An open-source llm reinforcement learning system at scale, 2025

    Qiying Yu, Zheng Zhang, Ruofei Zhu, Yufeng Yuan, Xiaochen Zuo, Yu Yue, Tiantian Fan, Gaohong Liu, Lingjun Liu, Xin Liu, Haibin Lin, Zhiqi Lin, Bole Ma, Guangming Sheng, Yuxuan Tong, Chi Zhang, Mofan Zhang, Wang Zhang, Hang Zhu, Jinhua Zhu, Jiaze Chen, Jiangjie Chen, Chengyi Wang, Hongli Yu, Weinan Dai, Yuxuan Song, Xiangpeng Wei, Hao Zhou, Jingjing Liu, W...

    2025

  5. [5]

    Process reinforcement through implicit rewards.arXiv preprint arXiv:2502.01456 , 2025

    Ganqu Cui, Lifan Yuan, Zefan Wang, Hanbin Wang, Wendi Li, Bingxiang He, Yuchen Fan, Tianyu Yu, Qixin Xu, Weize Chen, et al. Process reinforcement through implicit rewards.arXiv preprint arXiv:2502.01456 , 2025

    arXiv 2025

  6. [6]

    Scp-116k: A high-quality problem-solution dataset and a generalized pipeline for automated extraction in the higher education science domain, 2025

    Dakuan Lu, Xiaoyu Tan, Rui Xu, Tianchu Yao, Chao Qu, Wei Chu, Yinghui Xu, and Yuan Qi. Scp-116k: A high-quality problem-solution dataset and a generalized pipeline for automated extraction in the higher education science domain, 2025

    2025

  7. [7]

    Reasoning gym: Reasoning environments for reinforcement learning with verifiable rewards, 2025

    Zafir Stojanovski, Oliver Stanley, Joe Sharratt, Richard Jones, Abdulhakeem Adefioye, Jean Kaddour, and Andreas Köpf. Reasoning gym: Reasoning environments for reinforcement learning with verifiable rewards, 2025

    work page 2025

  8. [8]

    Nvidia, :, Bo Adler, Niket Agarwal, Ashwath Aithal, Dong H. Anh, Pallab Bhattacharya, Annika Brundyn, Jared Casper, Bryan Catanzaro, Sharon Clay, Jonathan Cohen, Sirshak Das, Ayush Dattagupta, Olivier Delalleau, Leon Derczynski, Yi Dong, Daniel Egert, Ellie Evans, Aleksander Ficek, Denys Fridman, Shaona Ghosh, Boris Ginsburg, Igor Gitman, Tomasz Grzegorze...

    work page 2024

Show all 33 references
  1. [9]

    Instruction-following evaluation for large language models, 2023

    Jeffrey Zhou, Tianjian Lu, Swaroop Mishra, Siddhartha Brahma, Sujoy Basu, Yi Luan, Denny Zhou, and Le Hou. Instruction-following evaluation for large language models, 2023

    2023

  2. [10]

    Deepseekmath: Pushing the limits of mathematical reasoning in open language models.arXiv preprint arXiv:2402.03300, 2024

    Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, YK Li, Y Wu, et al. Deepseekmath: Pushing the limits of mathematical reasoning in open language models.arXiv preprint arXiv:2402.03300, 2024

    2024 arXiv

  3. [11]

    Proximal policy optimization algorithms, 2017

    John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. Proximal policy optimization algorithms, 2017

    2017

  4. [12]

    Approximating KL Divergence

    John Schulman. Approximating KL Divergence

  5. [13]

    Deepcoder: A fully open-source 14b coder at o3-mini level

    Michael Luo, Sijun Tan, Roy Huang, Xiaoxiang Shi, Rachel Xin, Colin Cai, Ameen Patel, Alpay Ariyak, Qingyang Wu, Ce Zhang, Li Erran Li, Raluca Ada Popa, and Ion Stoica. Deepcoder: A fully open-source 14b coder at o3-mini level. https://pretty-radio-b75 .notion.site/DeepCoder-A...

    2025

  6. [14]

    Vapo: Efficient and reliable reinforcement learning for advanced reasoning tasks, 2025

    Yu Yue, Yufeng Yuan, Qiying Yu, Xiaochen Zuo, Ruofei Zhu, Wenyuan Xu, Jiaze Chen, Chengyi Wang, TianTian Fan, Zhengyin Du, Xiangpeng Wei, Xiangyu Yu, Gaohong Liu, Juncai Liu, Lingjun Liu, Haibin Lin, Zhiqi Lin, Bole Ma, Chi Zhang, Mofan Zhang, Wang Zhang, Hang Zhu, Ru Zhang, X...

    2025

  7. [15]

    Skywork open reasoner series.https://capricious-hydrogen-41c .notion.site/Skywork-Open-Reaonser- Series-1d0bc9ae823a80459b46c149e4f51680, 2025

    Jujie He, Jiacai Liu, Chris Yuhao Liu, Rui Yan, Chaojie Wang, Peng Cheng, Xiaoyu Zhang, Fuxiang Zhang, Jiacheng Xu, Wei Shen, Siyuan Li, Liang Zeng, Tianwen Wei, Cheng Cheng, Bo An, Yang Liu, and Yahui Zhou. Skywork open reasoner series.https://capricious-hydrogen-41c .notion....

    2025

  8. [16]

    Hybridflow: A flexible and efficient rlhf framework

    Guangming Sheng, Chi Zhang, Zilingfeng Ye, Xibin Wu, Wang Zhang, Ru Zhang, Yanghua Peng, Haibin Lin, and Chuan Wu. Hybridflow: A flexible and efficient rlhf framework. InProceedings of the Twentieth European Conference on Computer Systems , EuroSys ’25, page 1279–1297. ACM, March 2025

    2025

  9. [17]

    Decoupled weight decay regularization, 2019

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization, 2019

    2019

  10. [18]

    7b model and 8k examples: Emerging reasoning with reinforcement learning is both effective and efficient.https://hkust-nlp.notion.site/ simplerl-reason, 2025

    Weihao Zeng, Yuzhen Huang, Wei Liu, Keqing He, Qian Liu, Zejun Ma, and Junxian He. 7b model and 8k examples: Emerging reasoning with reinforcement learning is both effective and efficient.https://hkust-nlp.notion.site/ simplerl-reason, 2025. Notion Blog

    2025

  11. [19]

    American invitational mathematics examination - aime

    MAA. American invitational mathematics examination - aime. InAmerican Invitational Mathematics Examination - AIME 2024 , February 2024

    2024

  12. [20]

    American invitational mathematics examination - aime

    MAA. American invitational mathematics examination - aime. InAmerican Invitational Mathematics Examination - AIME 2025 , February 2025

    2025

  13. [21]

    American mathematics competition - amc

    MAA. American mathematics competition - amc. InAmerican Mathematics Competition - AMC

  14. [22]

    Measuring mathematical problem solving with the math dataset, 2021

    Dan Hendrycks, Collin Burns, Saurav Kadavath, Akul Arora, Steven Basart, Eric Tang, Dawn Song, and Jacob Steinhardt. Measuring mathematical problem solving with the math dataset, 2021

    2021

  15. [23]

    Solving quantitative reasoning problems with language models, 2022

    Aitor Lewkowycz, Anders Andreassen, David Dohan, Ethan Dyer, Henryk Michalewski, Vinay Ramasesh, Ambrose Slone, Cem Anil, Imanol Schlag, Theo Gutman-Solo, Yuhuai Wu, Behnam Neyshabur, Guy Gur-Ari, and Vedant Misra. Solving quantitative reasoning problems with language models, 2022

    2022

  16. [24]

    Olympiadbench: A challenging benchmark for promoting agi with olympiad-level bilingual multimodal scientific problems, 2024

    Chaoqun He, Renjie Luo, Yuzhuo Bai, Shengding Hu, Zhen Leng Thai, Junhao Shen, Jinyi Hu, Xu Han, Yujie Huang, Yuxiang Zhang, Jie Liu, Lei Qi, Zhiyuan Liu, and Maosong Sun. Olympiadbench: A challenging benchmark for promoting agi with olympiad-level bilingual multimodal scienti...

    2024

  17. [25]

    Measuring coding challenge competence with apps, 2021

    Dan Hendrycks, Steven Basart, Saurav Kadavath, Mantas Mazeika, Akul Arora, Ethan Guo, Collin Burns, Samir Puranik, Horace He, Dawn Song, and Jacob Steinhardt. Measuring coding challenge competence with apps, 2021

    2021

  18. [26]

    Mankowitz, Esme Sutherland Robson, Pushmeet Kohli, Nando de Freitas, Koray Kavukcuoglu, and Oriol Vinyals

    Yujia Li, David Choi, Junyoung Chung, Nate Kushman, Julian Schrittwieser, Rémi Leblond, Tom Eccles, James Keeling, Felix Gimeno, Agustin Dal Lago, Thomas Hubert, Peter Choy, Cyprien de Masson d’Autume, Igor Babuschkin, Xinyun Chen, Po-Sen Huang, Johannes Welbl, Sven Gowal, Ale...

    2022

  19. [27]

    Taco: Topics in algorithmic code generation dataset, 2023

    Rongao Li, Jie Fu, Bo-Wen Zhang, Tao Huang, Zhihong Sun, Chen Lyu, Guang Liu, Zhi Jin, and Ge Li. Taco: Topics in algorithmic code generation dataset, 2023

    2023

  20. [28]

    Is your code generated by chatGPT really correct? rigorous evaluation of large language models for code generation

    Jiawei Liu, Chunqiu Steven Xia, Yuyao Wang, and Lingming Zhang. Is your code generated by chatGPT really correct? rigorous evaluation of large language models for code generation. InThirty-seventh Conference on Neural Information Processing Systems, 2023

    2023

  21. [29]

    Livecodebench: Holistic and contamination free evaluation of large language models for code, 2024

    Naman Jain, King Han, Alex Gu, Wen-Ding Li, Fanjia Yan, Tianjun Zhang, Sida Wang, Armando Solar-Lezama, Koushik Sen, and Ion Stoica. Livecodebench: Holistic and contamination free evaluation of large language models for code, 2024

    2024

  22. [30]

    David Rein, Betty Li Hou, Asa Cooper Stickland, Jackson Petty, Richard Yuanzhe Pang, Julien Dirani, Julian Michael, and Samuel R. Bowman. Gpqa: A graduate-level google-proof q&a benchmark, 2023

    2023

  23. [31]

    Online difficulty filtering for reasoning oriented reinforcement learning, 2025

    Sanghwan Bae, Jiwoo Hong, Min Young Lee, Hanbyul Kim, JeongYeon Nam, and Donghyun Kwak. Online difficulty filtering for reasoning oriented reinforcement learning, 2025

    2025

  24. [32]

    Gonzalez, Hao Zhang, and Ion Stoica

    Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng, Lianmin Zheng, Cody Hao Yu, Joseph E. Gonzalez, Hao Zhang, and Ion Stoica. Efficient memory management for large language model serving with pagedattention. In Proceedings of the ACM SIGOPS 29th Symposium on Operating Systems...

    2023

  25. [33]

    The curious case of neural text degeneration, 2020

    Ari Holtzman, Jan Buys, Li Du, Maxwell Forbes, and Yejin Choi. The curious case of neural text degeneration, 2020. 14

    2020

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.