mlsanityData to 14 Oct 2026

Trade-offs in Large Reasoning Models: An Empirical Analysis of Deliberative and Adaptive Reasoning over Foundational Capabilities

arXiv 2503.17979 ↗First posted 23 Mar 2025cs.AI cs.CLcited by 20

On the map: LLM reasoning → Test-time compute and efficient thinking

Abstract (arXiv)

Recent advancements in Large Reasoning Models (LRMs), such as OpenAI's o1/o3 and DeepSeek-R1, have demonstrated remarkable performance in specialized reasoning tasks through human-like deliberative thinking and long chain-of-thought reasoning. However, our systematic evaluation across various model families (DeepSeek, Qwen, and LLaMA) and scales (7B to 32B) reveals that acquiring these deliberative reasoning capabilities significantly reduces the foundational capabilities of LRMs, including notable declines in helpfulness and harmlessness, alongside substantially increased inference costs. Importantly, we demonstrate that adaptive reasoning -- employing modes like Zero-Thinking, Less-Thinking, and Summary-Thinking -- can effectively alleviate these drawbacks. Our empirical insights underline the critical need for developing more versatile LRMs capable of dynamically allocating inference-time compute according to specific task characteristics.

Key facts. Cited by 20 papers in the map (data through 14 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 66 of its 80 references that are in the map. First posted on arXiv 23 Mar 2025.

Builds on

66 of 80 references.

Closest ancestors · 12

Same direction, newest first

  1. 2025Reasoning Effort and Problem Complexity: A Scaling Analysis in LLMs19 Mar
  2. 2025MetaScale: Test-Time Scaling with Evolving Meta-Thoughts17 Mar
  3. 2025L1: Controlling How Long A Reasoning Model Thinks With Reinforcement Learning6 Mar
  4. 2025The First Few Tokens Are All You Need: An Efficient and Effective Unsupervised Prefix Fine-Tuning Method for Reasoning Models4 Mar
  5. 2025The relationship between reasoning and performance in large language models--o3 (mini) thinks harder, not longer21 Feb
  6. 2025s1: Simple test-time scaling31 Jan
  7. 2025Thoughts Are All Over the Place: On the Underthinking of o1-Like LLMs30 Jan
  8. 2025O1-Pruner: Length-Harmonizing Fine-Tuning for O1-Like Reasoning Pruning22 Jan
  9. 2024Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs30 Dec
  10. 2024Learning How Hard to Think: Input-Adaptive Allocation of LM Computation7 Oct
  11. 2024Adaptive Inference-Time Compute: LLMs Can Predict if They Can Do Better, Even Mid-Generation3 Oct
  12. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters6 Aug

Other references · 54

  1. 2020Language Models are Few-Shot Learnerscited by 32,419
  2. 2022Training language models to follow instructions with human feedbackcited by 15,971
  3. 2024The Llama 3 Herd of Modelscited by 14,715
  4. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,626
  5. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,330
  6. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,944
  7. 2024Qwen2.5 Technical Reportcited by 4,942
  8. 2016Overcoming catastrophic forgetting in neural networkscited by 4,787
  9. 2024GPT-4o System Cardcited by 4,588
  10. 2024Gemma 2: Improving Open Language Models at a Practical Sizecited by 1,998
  11. 2024MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmarkcited by 1,848
  12. 2023Instruction-Following Evaluation for Large Language Modelscited by 1,228
38 more
  1. 2025Kimi k1.5: Scaling Reinforcement Learning with LLMscited by 1,210
  2. 2023WizardLM: Empowering large pre-trained language models to follow complex instructionscited by 1,125
  3. 2023A Comprehensive Survey of Continual Learning: Theory, Method and Applicationcited by 872
  4. 2023Enhancing Chat Language Models by Scaling High-quality Instructional Conversationscited by 839
  5. 2025LIMO: Less is More for Reasoningcited by 498
  6. 2023Safety-Tuned LLaMAs: Lessons From Improving the Safety of Large Language Models that Follow Instructionscited by 387
  7. 2024A StrongREJECT for Empty Jailbreakscited by 385
  8. 2025Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Modelscited by 384
  9. 2025From System 1 to System 2: A Survey of Reasoning Large Language Modelscited by 320
  10. 2023Orthogonal Subspace Learning for Language Model Continual Learningcited by 312
  11. 2025Towards Large Reasoning Models: A Survey of Reinforced Reasoning with Large Language Modelscited by 293
  12. 2025Training Language Models to Reason Efficientlycited by 247
  13. 2024Continual Learning of Large Language Models: A Comprehensive Surveycited by 234
  14. 2024WildTeaming at Scale: From In-the-Wild Jailbreaks to (Adversarially) Safer Language Modelscited by 223
  15. 2024Continual Learning with Pre-Trained Models: A Surveycited by 154
  16. 2024Continual Learning for Large Language Models: A Surveycited by 147
  17. 2025SafeChain: Safety of Language Models with Long Chain-of-Thought Reasoning Capabilitiescited by 136
  18. 2025LLM Post-Training: A Deep Dive into Reasoning Large Language Modelscited by 135
  19. 2025LLMs Can Easily Learn to Reason from Demonstrations Structure, not content, is what matters!cited by 132
  20. 2025Optimizing Test-Time Compute via Meta Reinforcement Fine-Tuningcited by 123
  21. 2025Safety Tax: Safety Alignment Makes Your Large Reasoning Models Less Reasonablecited by 122
  22. 2025Small Models Struggle to Learn from Strong Reasonerscited by 113
  23. 2025H-CoT: Hijacking the Chain-of-Thought Safety Reasoning Mechanism to Jailbreak Large Reasoning Models, Including OpenAI o1/o3, DeepSeek-R1, and Gemini 2.0 Flash Thinkingcited by 99
  24. 2025The Hidden Risks of Large Reasoning Models: A Safety Assessment of R1cited by 91
  25. 2025A Survey on Post-training of Large Language Modelscited by 54
  26. 2024SAPT: A Shared Attention Framework for Parameter-Efficient Continual Learning of Large Language Modelscited by 38
  27. 2025Unlocking General Long Chain-of-Thought Reasoning Capabilities of Large Language Models via Representation Engineeringcited by 36
  28. 2025Evaluating o1-Like LLMs: Unlocking Reasoning for Translation through Comprehensive Analysiscited by 26
  29. 2025Towards Understanding Distilled Reasoning Models: A Representational Approachcited by 26
  30. 2025Thinking Machines: A Survey of LLM based Reasoning Strategiescited by 25
  31. 2025o3-mini vs DeepSeek-R1: Which One is Safer?cited by 24
  32. 2025Cats Confuse Reasoning LLM: Query Agnostic Adversarial Triggers for Reasoning Modelscited by 23
  33. 2023ConPET: Continual Parameter-Efficient Tuning for Large Language Modelscited by 17
  34. 2025ThinkBench: Dynamic Out-of-Distribution Evaluation for Robust LLM Reasoningcited by 17
  35. 2025Reasoning Does Not Necessarily Improve Role-Playing Abilitycited by 14
  36. 2024Data Advisor: Dynamic Data Curation for Safety Alignment of Large Language Modelscited by 13
  37. 2025Marco-o1 v2: Towards Widening The Distillation Bottleneck for Reasoning Modelscited by 13
  38. 2025Challenges in Ensuring AI Safety in DeepSeek-R1 Models: The Shortcomings of Reinforcement Learning Strategiescited by 12

50 most cited of 54 All 54 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 14; 2026: 6.

20 papers.

Sort:

By area:

  1. 2025A Survey of Context Engineering for Large Language Modelscited by 126
  2. 2025100 Days After DeepSeek-R1: A Survey on Replication Studies and More Directions for Reasoning Language Modelscited by 30
  3. 2025A Comprehensive Survey on Trustworthiness in Reasoning with Large Language Modelscited by 25
  4. 2025Towards Concise and Adaptive Thinking in Large Reasoning Models: A Surveycited by 18
  5. 2025When Less Language is More: Language-Reasoning Disentanglement Makes LLMs Better Multilingual Reasonerscited by 16
  6. 2025Thinking in Character: Advancing Role-Playing Agents with Role-Aware Reasoningcited by 14
  7. 2025On Reasoning Strength Planning in Large Reasoning Modelscited by 13
  8. 2025Understanding the Thinking Process of Reasoning Models: A Perspective from Schoenfeld's Episode Theorycited by 9
  9. 2025BARREL: Boundary-Aware Reasoning for Factual and Reliable LRMscited by 8
  10. 2025Exploring and Exploiting the Inherent Efficiency within Large Reasoning Models for Self-Guided Efficiency Enhancementcited by 8
  11. 2025LoRA is All You Need for Safety Alignment of Reasoning LLMscited by 6
  12. 2026Towards a Mechanistic Understanding of Large Reasoning Models: A Survey of Training, Inference, and Failurescited by 6
8 more
  1. 2026STAR-S: Improving Safety Alignment through Self-Taught Reasoning on Safety Rulescited by 3
  2. 2025Compass-V2 Technical Reportcited by 2
  3. 2025FireScope: Wildfire Risk Raster Prediction with a Chain-of-Thought Oraclecited by 2
  4. 2025Understanding Chain-of-Thought Effectiveness in Code Generation: An Empirical and Information-Theoretic Analysiscited by 2
  5. 2026The Deliberative Deficit: An Empirical Critique of LLMs in Democratic Discoursecited by 2
  6. 2026From Trajectories to Prefixes: Reusing Teacher Trajectories via Replayed Prefixes and Online Continuationcited by 1
  7. 2026PPA-Plan: Proactive Pitfall Avoidance for Reliable Planning in Long-Context LLM Reasoningcited by 0
  8. 2026AdaThinking-E: One-Token Entropy Regulation for Adaptive Thinkingcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025A Survey of Efficient Reasoning for Large Reasoning Models: Language, Multimodality, and Beyond
  2. 2025Think or Not? Exploring Thinking Efficiency in Large Reasoning Models via an Information-Theoretic Lens
  3. 2025Efficient Inference for Large Reasoning Models: A Survey
  4. 2025THINK-Bench: Evaluating Thinking Efficiency and Chain-of-Thought Quality of Large Reasoning Models
  5. 2025Exploring the System 1 Thinking Capability of Large Reasoning Models
  6. 2025Don't Overthink It: A Survey of Efficient R1-style Large Reasoning Models
  7. 2025Large Reasoning Models are not thinking straight: on the unreliability of thinking trajectories
  8. 2025Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models