mlsanityData to 14 Oct 2026

Phi-4-reasoning Technical Report

arXiv 2504.21318 ↗First posted 30 Apr 2025cs.AI cs.CLcited by 151

On the map: LLM reasoning → Reasoning benchmarks and evaluation

Abstract (arXiv)

We introduce Phi-4-reasoning, a 14-billion parameter reasoning model that achieves strong performance on complex reasoning tasks. Trained via supervised fine-tuning of Phi-4 on carefully curated set of "teachable" prompts-selected for the right level of complexity and diversity-and reasoning demonstrations generated using o3-mini, Phi-4-reasoning generates detailed reasoning chains that effectively leverage inference-time compute. We further develop Phi-4-reasoning-plus, a variant enhanced through a short phase of outcome-based reinforcement learning that offers higher performance by generating longer reasoning traces. Across a wide range of reasoning tasks, both models outperform significantly larger open-weight models such as DeepSeek-R1-Distill-Llama-70B model and approach the performance levels of full DeepSeek-R1 model. Our comprehensive evaluations span benchmarks in math and scientific reasoning, coding, algorithmic problem solving, planning, and spatial understanding. Interestingly, we observe a non-trivial transfer of improvements to general-purpose benchmarks as well. In this report, we provide insights into our training data, our training methodologies, and our evaluations. We show that the benefit of careful data curation for supervised fine-tuning (SFT) extends to reasoning language models, and can be further amplified by reinforcement learning (RL). Finally, our evaluation points to opportunities for improving how we assess the performance and robustness of reasoning models.

Key facts. Cited by 151 papers in the map (data through 14 Oct 2026). Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning. Builds on 43 of its 56 references that are in the map. First posted on arXiv 30 Apr 2025.

Builds on

43 of 56 references.

Closest ancestors · 2

Same direction, newest first

  1. 2025A Sober Look at Progress in Language Model Reasoning: Pitfalls and Paths to Reproducibility9 Apr
  2. 2024Same Task, More Tokens: the Impact of Input Length on the Reasoning Performance of Large Language Models19 Feb

Other references · 41

  1. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 9,379
  2. 2020Measuring Massive Multitask Language Understandingcited by 7,963
  3. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,330
  4. 2024Qwen2.5 Technical Reportcited by 4,942
  5. 2021RoFormer: Enhanced Transformer with Rotary Position Embeddingcited by 4,703
  6. 2024GPT-4o System Cardcited by 4,588
  7. 2023GPQA: A Graduate-Level Google-Proof Q&A Benchmarkcited by 2,879
  8. 2024HybridFlow: A Flexible and Efficient RLHF Frameworkcited by 2,057
  9. 2024Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phonecited by 1,917
  10. 2016Concrete Problems in AI Safetycited by 1,642
  11. 2023Is Your Code Generated by ChatGPT Really Correct? Rigorous Evaluation of Large Language Models for Code Generationcited by 1,391
  12. 2023Instruction-Following Evaluation for Large Language Modelscited by 1,228
29 more
  1. 2025Kimi k1.5: Scaling Reinforcement Learning with LLMscited by 1,210
  2. 2023Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Promptingcited by 1,122
  3. 2022Scaling Laws for Reward Model Overoptimizationcited by 1,095
  4. 2023Extending Context Window of Large Language Models via Positional Interpolationcited by 740
  5. 2023Textbooks Are All You Needcited by 569
  6. 2022Language Models are Multilingual Chain-of-Thought Reasonerscited by 530
  7. 2024From Crowdsourced Data to High-Quality Benchmarks: Arena-Hard and BenchBuilder Pipelinecited by 500
  8. 2022ToxiGen: A Large-Scale Machine-Generated Dataset for Adversarial and Implicit Hate Speech Detectioncited by 422
  9. 2023Orca: Progressive Learning from Complex Explanation Traces of GPT-4cited by 379
  10. 2025Demystifying Long Chain-of-Thought Reasoning in LLMscited by 326
  11. 2025rStar-Math: Small LLMs Can Master Math Reasoning with Self-Evolved Deep Thinkingcited by 313
  12. 2024Omni-MATH: A Universal Olympiad Level Mathematic Benchmark For Large Language Modelscited by 284
  13. 2025Chain-of-Thought Reasoning In The Wild Is Not Always Faithfulcited by 170
  14. 2023Orca 2: Teaching Small Language Models How to Reasoncited by 159
  15. 2025LIMR: Less is More for RL Scalingcited by 139
  16. 2025SafeChain: Safety of Language Models with Long Chain-of-Thought Reasoning Capabilitiescited by 136
  17. 2024Is A Picture Worth A Thousand Words? Delving Into Spatial Reasoning for Vision Language Modelscited by 113
  18. 2024AgentInstruct: Toward Generative Teaching with Agentic Flowscited by 89
  19. 2025Reinforcement Learning for Reasoning in Small LLMs: What Works and What Doesn'tcited by 74
  20. 2025Proof or Bluff? Evaluating LLMs on 2025 USA Math Olympiadcited by 74
  21. 2018Towards Accountable AI: Hybrid Human-Machine Analyses for Characterizing System Failurecited by 49
  22. 2021Designing Disaggregated Evaluations of AI Systems: Choices, Considerations, and Tradeoffscited by 48
  23. 2025Inference-Time Scaling for Complex Tasks: Where We Stand and What Lies Aheadcited by 44
  24. 2025On the Emergence of Thinking in LLMs I: Searching for the Right Intuitioncited by 22
  25. 2025EXAONE Deep: Reasoning Enhanced Language Modelscited by 21
  26. 2024Eureka: Evaluating and Understanding Large Foundation Modelscited by 18
  27. 2023A Framework for Automated Measurement of Responsible AI Harms in Generative AI Applicationscited by 12
  28. 2023KITAB: Evaluating LLMs on Constraint Satisfaction for Information Retrievalcited by 11
  29. 2024Unearthing Skill-Level Insights for Understanding Trade-Offs of Foundation Modelscited by 9

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 72; 2026: 79.

151 papers.

Sort:

By area:

  1. 2025OpenThoughts: Data Recipes for Reasoning Modelscited by 304
  2. 2025A Survey of Reinforcement Learning for Large Reasoning Modelscited by 180
  3. 2026Beyond Benchmarks: MathArena as an Evaluation Platform for Mathematics with LLMscited by 137
  4. 2025The Invisible Leash: Why RLVR May or May Not Escape Its Origincited by 97
  5. 2025Don't Overthink it. Preferring Shorter Thinking Chains for Improved LLM Reasoningcited by 85
  6. 2025rStar2-Agent: Agentic Reasoning Technical Reportcited by 54
  7. 2025rStar-Coder: Scaling Competitive Code Reasoning with a Large-Scale Verified Datasetcited by 51
  8. 2025The ML.ENERGY Benchmark: Toward Automated Inference Energy Measurement and Optimizationcited by 33
  9. 2025EXAONE 4.0: Unified Large Language Models Integrating Non-reasoning and Reasoning Modescited by 32
  10. 2025BugPilot: Complex Bug Generation for Efficient Learning of SWE Skillscited by 30
  11. 2025FlashThink: An Early Exit Method For Efficient Reasoningcited by 28
  12. 2025Leaky Thoughts: Large Reasoning Models Are Not Private Thinkerscited by 28
38 more
  1. 2025EffiBench-X: A Multi-Language Benchmark for Measuring Efficiency of LLM-Generated Codecited by 27
  2. 2025System-1.5 Reasoning: Traversal in Language and Latent Spaces with Dynamic Shortcutscited by 27
  3. 2025Wait, Wait, Wait... Why Do Reasoning Models Loop?cited by 26
  4. 2025R3: Robust Rubric-Agnostic Reward Modelscited by 23
  5. 2025SwiReasoning: Switch-Thinking in Latent and Explicit for Pareto-Superior Reasoning LLMscited by 22
  6. 2025KTAE: A Model-Free Algorithm to Key-Tokens Advantage Estimation in Mathematical Reasoningcited by 19
  7. 2025VoiceAgentBench: Are Voice Assistants ready for agentic tasks?cited by 19
  8. 2025Don't Pass@k: A Bayesian Framework for Large Language Model Evaluationcited by 15
  9. 2025T1: A Tool-Oriented Conversational Dataset for Multi-Turn Agentic Planningcited by 14
  10. 2025Value-Guided Search for Efficient Chain-of-Thought Reasoningcited by 13
  11. 2025MIRAGE: A Benchmark for Multimodal Information-Seeking and Reasoning in Agricultural Expert-Guided Conversationscited by 13
  12. 2026Falcon-H1R: Pushing the Reasoning Frontiers with a Hybrid Model for Efficient Test-Time Scalingcited by 13
  13. 2026Which Reasoning Trajectories Teach Students to Reason Better? A Simple Metric of Informative Alignmentcited by 13
  14. 2025SynthRL: Scaling Visual Reasoning with Verifiable Data Synthesiscited by 12
  15. 2025ThinkQE: Query Expansion via an Evolving Thinking Processcited by 12
  16. 2025ReAcTree: Hierarchical LLM Agent Trees with Control Flow for Long-Horizon Task Planningcited by 12
  17. 2025PrivacyBench: A Conversational Benchmark for Evaluating Privacy in Personalized AIcited by 12
  18. 2025K2-Think: A Parameter-Efficient Reasoning Systemcited by 11
  19. 2025Refusal Falls off a Cliff: How Safety Alignment Fails in Reasoning?cited by 11
  20. 2026The Last AI Built by Humans: Toward Genuine Recursive Self-Improvementcited by 11
  21. 2025ReProbe: Efficient Test-Time Scaling of Multi-Step Reasoning by Probing Internal States of Large Language Modelscited by 10
  22. 2025Understanding and Steering the Cognitive Behaviors of Reasoning Models at Test-Timecited by 10
  23. 2026Hidden States as Early Signals: Step-level Trace Evaluation and Pruning for Efficient Test-Time Scalingcited by 10
  24. 2025First Finish Search: Efficient Test-Time Scaling in Large Language Modelscited by 9
  25. 2025SATA-BENCH: Select All That Apply Benchmark for Multiple Choice Questionscited by 9
  26. 2025Tracing the Traces: Latent Temporal Signals for Efficient and Accurate Reasoningcited by 9
  27. 2026Do Not Waste Your Rollouts: Recycling Search Experience for Efficient Test-Time Scalingcited by 9
  28. 2025ToTRL: Unlock LLM Tree-of-Thoughts Reasoning Potential through Puzzles Solvingcited by 8
  29. 2025Interpretable Traces, Unexpected Outcomes: Investigating the Disconnect in Trace-Based Knowledge Distillationcited by 8
  30. 2025Med-U1: Incentivizing Unified Medical Reasoning in LLMs via Large-scale Reinforcement Learningcited by 8
  31. 2025Which LLMs Get the Joke? Probing Non-STEM Reasoning Abilities with HumorBenchcited by 8
  32. 2025Tiny Model, Big Logic: Diversity-Driven Optimization Elicits Large-Model Reasoning Ability in VibeThinker-1.5Bcited by 8
  33. 2026VLM Judges Can Rank but Cannot Score: Task-Dependent Uncertainty in Multimodal Evaluationcited by 8
  34. 2025Longer Context, Deeper Thinking: Uncovering the Role of Long-Context Ability in Reasoningcited by 7
  35. 2025AlgoTrace: Algorithmic Primitives and Compositional Geometry of Reasoning in Language Modelscited by 7
  36. 2026Llama-3.1-FoundationAI-SecurityLLM-Reasoning-8B Technical Reportcited by 7
  37. 2026SeLaR: Selective Latent Reasoning in Large Language Modelscited by 7
  38. 2025DEBATE, TRAIN, EVOLVE: Self Evolution of Language Model Reasoningcited by 6

50 most cited of 151 All 151 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025Phi-4-Mini-Reasoning: Exploring the Limits of Small Reasoning Language Models in Math
  2. 2024Phi-4 Technical Report
  3. 2026Phi-4-reasoning-vision-15B Technical Report
  4. 2025DeepMath-103K: A Large-Scale, Challenging, Decontaminated, and Verifiable Mathematical Dataset for Advancing Reasoning
  5. 2023Evaluating the Logical Reasoning Ability of ChatGPT and GPT-4
  6. 2025Are DeepSeek R1 And Other Reasoning Models More Faithful?
  7. 2025Evaluating Mathematical Reasoning Across Large Language Models: A Fine-Grained Approach
  8. 2025AceReason-Nemotron: Advancing Math and Code Reasoning through Reinforcement Learning