The DOI matches its DataCite record; 68 of the 76 references it cites were found in a registry or an index, and 8 are not checked yet; 2 of the 17 sources its claims cite were checked for support.
DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
Open accessPreprint — published as 10.1038/s41586-025-09422-z · Paper page
Abstract
General reasoning represents a long-standing and formidable challenge in artificial intelligence. Recent breakthroughs, exemplified by large language models (LLMs) and chain-of-thought prompting, have achieved considerable success on foundational reasoning tasks. However, this success is heavily contingent upon extensive human-annotated demonstrations, and models' capabilities are still insufficient for more complex problems. Here we show that the reasoning abilities of LLMs can be incentivized through pure reinforcement learning (RL), obviating the need for human-labeled reasoning trajectories. The proposed RL framework facilitates the emergent development of advanced reasoning patterns, such as self-reflection, verification, and dynamic strategy adaptation. Consequently, the trained model achieves superior performance on verifiable tasks such as mathematics, coding competitions, and STEM fields, surpassing its counterparts trained via conventional supervised learning on human demonstrations. Moreover, the emergent reasoning patterns exhibited by these large-scale models can be systematically harnessed to guide and enhance the reasoning capabilities of smaller models.
Verified 2026-10-05 · 8 references not checked yet
Overview
CoChat Summary
TL;DR
DeepSeek-R1 shows that large-scale reinforcement learning can incentivize reasoning in LLMs without human-labeled reasoning trajectories, producing emergent self-reflection, verification, and longer chains of thought. A multi-stage pipeline then yields DeepSeek-R1, which reaches frontier reasoning performance and can be distilled into smaller models.
Key findings
- Pure RL on DeepSeek-V3-Base, without SFT, produced DeepSeek-R1-Zero: AIME 2024 pass@1 rose from 15.6% to 77.9%, and self-consistency decoding reached 86.7%.
- DeepSeek-R1 achieved AIME 2024 pass@1 79.8, MATH-500 97.3, Codeforces rating 2029 (96.3 percentile), LiveCodeBench 65.9, MMLU 90.8, AlpacaEval 2.0 87.6, and ArenaHard 92.3.
- Distillation transferred reasoning to smaller models: DeepSeek-R1-Distill-Qwen-1.5B reached AIME 2024 28.9 and MATH 83.9; DeepSeek-R1-Distill-Qwen-32B reached AIME 2024 72.6, MATH 94.3, GPQA Diamond 62.1, LiveCodeBench 57.2, and Codeforces 1691.
- During RL, reflective-word frequency rose 5- to 7-fold, and the model showed an “aha moment” with increased use of “wait,” indicating emergent reflective and verification behaviors.
Methods
- DeepSeek-R1-Zero was trained from DeepSeek-V3-Base using Group Relative Policy Optimization (GRPO) with rule-based accuracy and format rewards, bypassing SFT; DeepSeek-R1 used a multi-stage pipeline of cold-start data, RL, rejection sampling plus SFT, and a second RL stage with model-based rewards.
- RL data included 26K math, 17K code, 22K STEM, 15K logic, and 66K general prompts; SFT data totaled 804,745 samples, and distillation used 800K samples.
- Evaluation covered MMLU, MMLU-Redux, MMLU-Pro, C-Eval, CMMLU, IFEval, FRAMES, GPQA Diamond, SimpleQA, C-SimpleQA, SWE-Bench Verified, Aider, LiveCodeBench, Codeforces, CNMO 2024, and AIME 2024, using pass@k with temperature 0.6 and top-p 0.95, plus n-gram decontamination.
Limitations
- Paper-stated: structural output and tool use remain suboptimal; token efficiency suffers from overthinking on simple questions; language mixing persists outside Chinese and English; few-shot prompting degrades performance; software engineering tasks received limited large-scale RL due long evaluation times; reward hacking remains a challenge when reliable reward models are unavailable; PRM and MCTS attempts were unsuccessful; effectiveness depends on sufficiently large base checkpoints; SFT data are mostly single-turn, limiting multi-turn conversation.
- Paper-stated safety/ethical: R1 can be jailbroken to generate dangerous content, public models are vulnerable to safety-compromising fine-tuning, and without risk control its in-house unsafe rate was 25.2% and jailbreak unsafe rate 85.9%, though with risk control these fell to 8.5% and 4.3%.
- Evident from the text: n-gram decontamination cannot prevent paraphrase contamination, so benchmarks released before 2024 may still be contaminated; evaluation also relies on LLM judges and third-party/HELM results for some safety comparisons.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
2 Valid · 14 Not valid · 8 Not checked · 243 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Third, once a model-based PRM is introduced, it inevitably leads to reward hacking (Gao et al., 2022)”
What the sources say
- Partly supportedcites Gao et al., 2022 sourceExperimental — accuracy not yet measured
Our copy supports the general mechanism only: the Abstract (p. 1) and Introduction (p. 1) state that because a reward model is an imperfect proxy, 'optimizing its value too much can hinder ground truth performance', and Related Work (p. 11) notes overoptimization of RMs 'can be viewed as a special case of specification gaming (also known as reward hacking)'. Two parts of the claim are not supported: the paper never studies process reward models or step-level/model-based PRMs — its setup (Methodology, p. 2) is an outcome-level preference RM in the InstructGPT environment optimized by RL or best-of-n — and it does not treat reward hacking as inevitable upon introduction of such a model. Section 4.2.1 (p. 9) explicitly says that if only regressional Goodhart were present 'the gold reward must always increase monotonically with the proxy reward', and the harm is tied to the amount of optimization pressure, so the citing sentence's unconditional 'inevitably leads to reward hacking' overstates a hedged, quantity-dependent finding. (LLM, conf=0.78)
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“In conclusion, while PRM demonstrates a good ability to rerank the top-N responses generated by the model or assist in guided search (Snell et al., 2024), its advantages are limited compared to the additional computational overhead it introduces during the large-scale reinforcement learning process in our experiments.”
What the sources say
- Partly supportedcites Snell et al., 2024 sourceExperimental — accuracy not yet measured
Our copy supports the capability the citation is attached to: Figure 2 (Appendix J) describes Best-of-N selecting the best of N full answers "according to the PRM final score" and beam/lookahead search retaining steps "according to the PRM", and Section 5.2 plus Appendix F/Figure 14 report the PRM consistently and greatly outperforming the ORM. What our copy does not support is the quoted conclusion that the PRM's advantages are limited relative to the extra computational overhead in a large-scale reinforcement learning process: Snell et al. train their PRM by Monte-Carlo-rollout supervision rather than by RL (Section 5.1, Appendix D), their only cost remarks are that lookahead "needs more compute" and that per-step PRM labels are costly to generate (Appendix J), and their overall result is that compute-optimal allocation against the PRM beats a best-of-N baseline "while only using about 4x less computation". That limiting clause is the citing authors' own experimental finding ("in our experiments"), and our copy says nothing about it. (LLM, conf=0.72)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In contrast, DeepSeek-R1 Dev2 demonstrates marked performance enhancements on benchmarks that require advanced reasoning skills, including those focused on code generation, mathematical problem solving, and STEM-related tasks.”
boostedstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“This performance significantly surpasses the average performance across all human competitors.”
boosted- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Although DeepSeek-R1-Zero demonstrates excellent reasoning capabilities, it faces challenges such as poor readability and language mixing, occasionally combining English and Chinese within a single chain-of-thought response.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
135 more claims
Create a free account to see all 267 claims and to verify the 1 reference still open.
- Probably not a substantial claim. 63
- It may state two separate claims as one. 83
- The quote only partly states this claim. 3
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Following this direction, researchers have proposed several methods for synthesizing instructions that elicit long-CoT reasoning.”
From A Survey of Large Language Models (2026) · cites this paper as [29]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Additionally, prompting strategies can be used to encourage the model to produce detailed reasoning processes with self-reflection and self-verification [29].”
From A Survey of Large Language Models (2026) · cites this paper as [29]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“For example, on the widely used MMLU benchmark [325], state-of-the-art models now exceed human-level performance after large scale training, with DeepSeek-R1 achieving over 90 % accuracy”
From A Survey of Large Language Models (2026) · cites this paper as [29]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“It should be noted that verifiability has long been central to scaling AI progress: self-supervised learning employs next-token prediction as a provable training objective [61]; instruction-tuning leverages LLM-generated instruction-response pairs as reliable signals [62]; and reinforcement learning uses verifiable rewards [55] for alignment.”
From When AI Co-Scientists Fail: SPOT-a Benchmark for Automated Verification of Scientific Research (2025) · cites this paper as [55]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“More recently RL has been extensively used for improving code and math performance in verifiable domains [34,64,89,134,204]”
From RLHF Deciphered: A Critical Analysis of Reinforcement Learning from Human Feedback for LLMs (2026) · cites this paper as [34]
Experimental — accuracy not yet measured
44 more claims about this paper
Create a free account to see every claim other papers make about this one.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ B.2. Reward Model Prompt
“| Code | 17K | Algorithm and Bug Fixing | Code Solution |”
§ B.3.1. RL Data
“Coding dataset includes 17k algorithm competition questions, along with 8k bug fixing problems.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ B.4.4. Training Cost
“completing the process in about 4 days, or roughly 80 hours.”
§ B.6. Ablation Study of Language Consistency Reward
“| in H800 GPU Hours | 101K | 5K | 41K | 147K |”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ B.2. Reward Model Prompt
“| General | 66K | Helpfulness/Harmlessness | Ranked Responses |”
§ B.3.1. RL Data
“General dataset consists of 66k questions designed to assess helpfulness, spanning various categories such as creative writing, editing, factual question answering, and role-playing. Additionally, the dataset includes 12,000 questions focused on evaluating harmlessness.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ D.3.5. Robustness against Jailbreaking
“the two reasoning models in our experiments DeepSeek-R1 and o1(2024-12-17) - rely more heavily on the risk control system for security checks, resulting in considerably higher overall rejection rates (79.8% and 87.3% respectively).”
§ D.3.5. Robustness against Jailbreaking
“| + risk control system | 8.5 | 4.3 | -4.2 | 27.3 | 87.3 | +60.0 |”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ D.3.5. Robustness against Jailbreaking
“All tested models exhibited significantly increased rates of unsafe responses and rejections, along with decreased safety rates when facing jailbreak attacks.”
§ D.3.5. Robustness against Jailbreaking
“| + risk control system | 8.5 | 4.3 | -4.2 | 27.3 | 87.3 | +60.0 |”
Citations
Cites
The publisher hasn’t deposited this paper’s reference list.
Part of this paper’s reference list could not be split into single entries.
Cited by
A Survey of Large Language Models
Zhao, W. X., Zhou, K., Li, J., Tang, T., Dong, Z., Hou, Y., Zhang, B., Min, Y., Zhang, J., Liu, P., et al. (2026) · Frontiers of Computer Science · doi:10.1007/s11704-026-60308-3 · Cited by 2,093
Self-citationFrom Generation to Judgment: Opportunities and Challenges of LLM-as-a-judge
Li, D., Jiang, B., Huang, L., Beigi, A., Zhao, C., Tan, Z., Bhattacharjee, A., Jiang, Y., Chen, C., Wu, T., et al. (2024) · arXiv · doi:10.48550/arxiv.2411.16594 · Cited by 271
Self-citationThe landscape of agentic reinforcement learning for LLMs: a survey
Zhang, G., Geng, H., Yu, X., Yin, Z., Zhang, Z., Tan, Z., Zhou, H., Li, Z., Xue, X., Li, Y., et al. (2025) · arXiv · doi:10.48550/arxiv.2509.02547 · Cited by 229
Self-citationBenchmark evaluation of DeepSeek large language models in clinical decision-making
Sandmann, S., Hegselmann, S., Fujarski, M., Bickmann, L., Wild, B., Eils, R., & Varghese, J. (2025) · Nature Medicine · doi:10.1038/s41591-025-03727-2 · Cited by 233
“13).”
Large Language Models for Information Retrieval: A Survey
Zhu, Y., Yuan, H., Wang, S., Liu, J., Liu, W., Deng, C., Chen, H., Liu, Z., Dou, Z., & Wen, J.-R. (2026) · ACM Transactions on Information Systems · doi:10.1145/3748304 · Cited by 225
“In the following section, we will introduce these three methods in detail.”
Self-citation
Metadata
- Authors
- DeepSeek-AI, Guo, D., Yang, D., Zhang, H., Song, J., Wang, P., Zhu, Q., Xu, R., Zhang, R., Ma, S., Bi, X., Zhang, X., Yu, X., Wu, Y., Wu, Z. F., Gou, Z., Shao, Z., Li, Z., Gao, Z., Liu, A., Xue, B., Wang, B., Wu, B., Feng, B., Lu, C., Zhao, C., Deng, C., Zhang, C., Ruan, C., Dai, D., Chen, D., Ji, D., Li, E., Lin, F., Dai, F., Luo, F., Hao, G., Chen, G., Li, G., Zhang, H., Bao, H., Xu, H., Wang, H., Ding, H., Xin, H., Gao, H., Qu, H., Li, H., Guo, J., Li, J., Wang, J., Chen, J., Yuan, J., Qiu, J., Li, J., Cai, J. L., Ni, J., Liang, J., Chen, J., Dong, K., Hu, K., Gao, K., Guan, K., Huang, K., Yu, K., Wang, L., Zhang, L., Zhao, L., Wang, L., Zhang, L., Xu, L., Xia, L., Zhang, M., Zhang, M., Tang, M., Li, M., Wang, M., Li, M., Tian, N., Huang, P., Zhang, P., Wang, Q., Chen, Q., Du, Q., Ge, R., Zhang, R., Pan, R., Wang, R., Chen, R. J., Jin, R. L., Chen, R., Lu, S., Zhou, S., Chen, S., Ye, S., Wang, S., Yu, S., Zhou, S., Pan, S., Li, S. S., Zhou, S., Wu, S., Ye, S., Yun, T., Pei, T., Sun, T., Wang, T., Zeng, W., Zhao, W., Liu, W., Liang, W., Gao, W., Yu, W., Zhang, W., Xiao, W. L., An, W., Liu, X., Wang, X., Chen, X., Nie, X., Cheng, X., Liu, X., Xie, X., Liu, X., Yang, X., Li, X., Su, X., Lin, X., Li, X. Q., Jin, X., Shen, X., Chen, X., Sun, X., Wang, X., Song, X., Zhou, X., Wang, X., Shan, X., Li, Y. K., Wang, Y. Q., Wei, Y. X., Zhang, Y., Xu, Y., Li, Y., Zhao, Y., Sun, Y., Wang, Y., Yu, Y., Zhang, Y., Shi, Y., Xiong, Y., He, Y., Piao, Y., Wang, Y., Tan, Y., Ma, Y., Liu, Y., Guo, Y., Ou, Y., Wang, Y., Gong, Y., Zou, Y., He, Y., Xiong, Y., Luo, Y., You, Y., Liu, Y., Zhou, Y., Zhu, Y. X., Xu, Y., Huang, Y., Li, Y., Zheng, Y., Zhu, Y., Ma, Y., Tang, Y., Zha, Y., Yan, Y., Ren, Z. Z., Ren, Z., Sha, Z., Fu, Z., Xu, Z., Xie, Z., Zhang, Z., Hao, Z., Ma, Z., Yan, Z., Wu, Z., Gu, Z., Zhu, Z., Liu, Z., Li, Z., Xie, Z., Song, Z., Pan, Z., Huang, Z., Xu, Z., Zhang, Z., & Zhang, Z.
- Affiliations
- DeepSeek (China) (CN)
- University of Science and Technology of China (CN)
- Peking University (CN)
- Corresponding author
- Liang, Wenfeng
- ORCID
- Guo, Daya
- Zhang, Haowei
- Song, Junxiao
- Wang, Peiyi
- Zhu, Qihao
- Zhang, Ruoyu
- Ma, Shirong
- Zhang, Xiaokang
- Yu, Xingkai
- Shao, Zhihong
- Li, Zhuoshu
- Gao, Ziyi
- Liu, Aixin
- Xue, Bing
- Lu, Chengda
- Zhao, Chenggang
- Deng, Chengqi
- Ruan, Chong
- Dai, Damai
- Chen, Deli
- Li, Guowei
- Xu, Hanwei
- Gao, Huazuo
- Li, Hui
- Guo, Jianzhong
- Li, Jiashi
- Chen, Jingchang
- Qiu, Junjie
- Li, Junlong
- Liang, Jian
- Hu, Kai
- Huang, Kexin
- Wang, Lean
- Zhao, Liang
- Wang, Litong
- Zhang, Liyue
- Xu, Lei
- Zhang, Mingchuan
- Zhang, Minghua
- Li, Meng
- Li, Mingming
- Huang, Panpan
- Chen, Qinyu
- Zhang, Ruisong
- Chen, Ruyi
- Zhou, Shangyan
- Wang, Shiyu
- Li, S. S.
- Zhou, Shuang
- Sun, Tianyu
- Liang, Wenfeng
- Gao, Wenjun
- Zhang, Wentao
- Liu, Xiaodong
- Wang, Xiaohan
- Chen, Xiaokang
- Cheng, Xin
- Liu, Xin
- Xie, Xin
- Liu, Xingchao
- Yang, Xinyu
- Li, Xinyuan
- Chen, Xiaosha
- Li, Y. K.
- Wang, Y. Q.
- Zhang, Yang
- Li, Yao
- Zhao, Yao
- Wang, Yaohui
- Yu, Yi
- Zhang, Yichao
- He, Ying
- Wang, Yisong
- Ma, Yiyang
- Liu, Yiyuan
- Guo, Yongqiang
- Luo, Yuxiang
- Liu, Yuxuan
- Zhou, Yuyang
- Zhu, Y. X.
- Li, Yaohui
- Tang, Ying
- Zha, Yukun
- Ren, Z. Z.
- Ren, Zehui
- Hao, Zhewen
- Ma, Zhicheng
- Wu, Zhiyu
- Gu, Zihui
- Liu, Zijun
- Li, Zilin
- Xie, Ziwei
- Song, Ziyang
- Pan, Zizheng
- Huang, Zhen
- Xu, Zhipeng
- Zhang, Zhongyu
- Zhang, Zhen
- Venue
- arXiv
- Publisher
- arXiv
- Published
- 2025
- Volume
- 645
- Issue
- 8081
- Pages
- 633-638
- ISSN
- 0028-0836 (ISSN-L)
- arXiv
- 2501.12948
- PMC
- PMC12443585
- OpenAlex
- W4414281281
- Semantic Scholar
- 2eed1fad9bbf887d4395de40f20144c4fafefd7f
- Cited by
- 18,918
- References
- 0
- Citations by year
- 2026: 11341 · 2025: 7464
- FWCI
- 4113.79
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- Locations
- Nature (publishedVersion, cc-by)
- Europe PMC (PubMed Central) (publishedVersion, cc-by)
- PubMed (publishedVersion)
- PubMed Central (publishedVersion, cc-by)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
- Physical Sciences › Computer Science › Artificial Intelligence › AI-based Problem Solving and Planning
- Keywords
- DeepSeek-R1, large language models, reasoning, reinforcement learning, chain-of-thought prompting, self-reflection, verification, strategic adaptation, mathematical reasoning
- MeSH terms
- Large Language Models, Artificial Intelligence, Humans, Motivation, Reinforcement, Psychology, Thinking
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-09-28, under version 3 of the claim definition (prompt extract-v3).
A second model labelled each claim on 2026-09-28 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (qwen3.8-omni-flash).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning. https://app.cochat.ai/papers/10.48550/arxiv.2501.12948
- The paper
- DeepSeek-AI, Guo, D., Yang, D., Zhang, H., Song, J., Wang, P., Zhu, Q., Xu, R., Zhang, R., Ma, S., Bi, X., Zhang, X., Yu, X., Wu, Y., Wu, Z. F., Gou, Z., Shao, Z., Li, Z., Gao, Z., . . . Zhang, Z. (2025). DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning. arXiv. https://doi.org/10.48550/arxiv.2501.12948
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.