Pith. sign in

REVIEW 15 cited by

Light-R1: Curriculum SFT, DPO and RL for Long COT from Scratch and Beyond

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2503.10460 v4 pith:2PXFC44X submitted 2025-03-13 cs.CL cs.LG

classification cs.CLcs.LG
keywords modelsdatatrainingcurriculumlight-r1reasoninglongapproach
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

This paper introduces Light-R1, an open-source suite for training long reasoning models using reproducible and cost-effective methodology. Given the proprietary nature of data used in the DeepSeek-R1 series, we develop an alternative approach leveraging exclusively public data and models. Our curriculum training progressively increases data difficulty, combined with multi-staged post-training. Our Light-R1-32B model, trained from Qwen2.5-32B-Instruct, outperforms DeepSeek-R1-Distill-Qwen-32B in math reasoning. Experimental results show that this curriculum approach becomes more effective when distinct, diverse datasets are available for different training stages: fine-tuning DeepSeek-R1-Distilled models (pre-tuned by DeepSeek team on proprietary data) with 3,000 challenging examples from our curriculum dataset yielded state-of-the-art 7B and 14B models, while the 32B model, Light-R1-32B-DS performed comparably to QwQ-32B and DeepSeek-R1. Furthermore, we extend our work by applying GRPO on long reasoning models. Our final Light-R1-14B-DS achieves SOTA performance among 14B models in math, with AIME24 & 25 scores of 74.0 and 60.2 respectively, surpassing many 32B models and DeepSeek-R1-Distill-Llama-70B. Despite math-focused training, Light-R1-14B-DS demonstrates strong cross-domain generalization. Light-R1 represents a significant advancement in making sophisticated reasoning models more accessible and implementable in real-world applications. Our models, training data and code have been made available at https://github.com/Qihoo360/Light-R1.

Discussion (0). Sign in to comment.

Forward citations

Cited by 15 Pith papers

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

  1. Sticker-TTS: Learn to Utilize Historical Experience with a Sticker-driven Test-Time Scaling Framework

    cs.AI 2025-09 conditional novelty 6.0 of 10

    Sticker-TTS uses three collaborating models to distill and reuse compact summaries (stickers) of past reasoning attempts, improving math benchmark accuracy at a claimed equivalent inference cost.

  2. Can One Domain Help Others? A Data-Centric Study on Multi-Domain Reasoning via Reinforcement Learning

    cs.AI 2025-07 conditional novelty 6.0 of 10

    Training a Qwen2.5-7B model with GRPO on math and puzzle data improves both domains, code transfer depends on the starting model, and template or reward mismatches sharply hurt performance.

  3. Squeeze the Soaked Sponge: Efficient Off-policy Reinforcement Finetuning for Large Language Model

    cs.LG 2025-07 conditional novelty 6.0 of 10

    ReMix cuts reinforcement finetuning rollout volume by 30x to 450x on math reasoning by mixing historical and on-policy data with a convex KL constraint and a mid-training switch to on-policy updates.

  4. MiCoTA: Bridging the Learnability Gap with Intermediate CoT and Teacher Assistants

    cs.CL 2025-07 conditional novelty 6.0 of 10

    Training small language models on intermediate-length reasoning chains from a merged mid-sized teacher assistant improves their math reasoning scores over direct distillation from a large teacher.

  5. A Model Can Help Itself: Reward-Free Self-Training for LLM Reasoning

    cs.LG 2025-10 unverdicted novelty 5.0 of 10

    SePT enables LLMs to improve math reasoning on multiple benchmarks by iteratively training on their own low-temperature generated responses using an online data refresh mechanism.

  6. Domain-Aware RAG: MoL-Enhanced RL for Efficient Training and Scalable Retrieval

    cs.CL 2025-09 conditional novelty 5.0 of 10

    A two-stage RAG training pipeline, MoL continual pre-training plus GRPO reinforcement learning with single-passage training and multi-passage inference, reports state-of-the-art retrieval recall on NFCORPUS and SCIFAC...

  7. Uncertainty Under the Curve: A Sequence-Level Entropy Area Metric for Reasoning LLM

    cs.AI 2025-08 conditional novelty 5.0 of 10

    Entropy Area Score sums token-level predictive entropy across a reasoning sequence, correlates with answer entropy, and selects SFT training data better than Pass Rate filtering in limited AIME experiments.

  8. ThinkDial: An Open Recipe for Controlling Reasoning Effort in Large Language Models

    cs.CL 2025-08 conditional novelty 5.0 of 10

    An open training recipe that teaches a 32B LLM to reason in three adjustable modes, cutting token use by half or three quarters with small accuracy losses.

  9. REST: Stress Testing Large Reasoning Models by Asking Multiple Problems at Once

    cs.CL 2025-07 conditional novelty 5.0 of 10

    Asking a reasoning model several problems at once reveals large accuracy drops and exposes differences that single-question benchmarks miss.

  10. Enhancing Large Language Models through Structured Reasoning

    cs.CL 2025-06 conditional novelty 5.0 of 10

    Structured reasoning tags plus a max-flow reward let a 1.5B model match the math accuracy of models trained for far longer, but the gains are within statistical noise.

  11. AceReason-Nemotron 1.1: Advancing Math and Code Reasoning through SFT and RL Synergy

    cs.CL 2025-06 conditional novelty 5.0 of 10

    A 7B reasoning model trained with carefully balanced SFT and RL beats prior small models on math and code benchmarks, with the paper documenting scaling and temperature heuristics.

  12. SwS: Self-aware Weakness-driven Problem Synthesis in Reinforcement Learning for LLM Reasoning

    cs.LG 2025-06 conditional novelty 5.0 of 10

    SwS uses failures during RL training to synthesize targeted math problems, improving reasoning accuracy on eight benchmarks.

  13. A Practical Two-Stage Recipe for Mathematical LLMs: Maximizing Accuracy with SFT and Efficiency with Reinforcement Learning

    cs.LG 2025-07 conditional novelty 4.0 of 10

    Ten-epoch SFT then GRPO raises AIME/MATH-500 accuracy while reducing solution length, achieving 8th place on AIMO's private set.

  14. Schema-R1: A reasoning training approach for schema linking in Text-to-SQL Task

    cs.AI 2025-06 conditional novelty 4.0 of 10

    Schema-R1 combines cold-start SFT on 200 CoT samples with GRPO rule-based RL, reporting table and column filter accuracy gains of 10 percentage points or more over a fine-tuned baseline on Spider-dev.

  15. Towards Concise and Adaptive Thinking in Large Reasoning Models: A Survey

    cs.AI 2025-07 conditional novelty 3.0 of 10

    A comprehensive review that categorizes methods for shortening and adaptively triggering chain-of-thought reasoning in large language models.

Pith tools