mlsanityData to 7 Oct 2026

Self-Discover: Large Language Models Self-Compose Reasoning Structures

arXiv 2402.03620 ↗First posted 6 Feb 2024cs.AI cs.CLcited by 108

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

Abstract (arXiv)

We introduce SELF-DISCOVER, a general framework for LLMs to self-discover the task-intrinsic reasoning structures to tackle complex reasoning problems that are challenging for typical prompting methods. Core to the framework is a self-discovery process where LLMs select multiple atomic reasoning modules such as critical thinking and step-by-step thinking, and compose them into an explicit reasoning structure for LLMs to follow during decoding. SELF-DISCOVER substantially improves GPT-4 and PaLM 2's performance on challenging reasoning benchmarks such as BigBench-Hard, grounded agent reasoning, and MATH, by as much as 32% compared to Chain of Thought (CoT). Furthermore, SELF-DISCOVER outperforms inference-intensive methods such as CoT-Self-Consistency by more than 20%, while requiring 10-40x fewer inference compute. Finally, we show that the self-discovered reasoning structures are universally applicable across model families: from PaLM 2-L to GPT-4, and from GPT-4 to Llama2, and share commonalities with human reasoning patterns.

Key facts. Cited by 108 papers in the map (data through 7 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 41 of its 52 references that are in the map. First posted on arXiv 6 Feb 2024.

Builds on

41 of 52 references.

Closest ancestors · 1

Same direction, newest first

  1. 2023Tree of Thoughts: Deliberate Problem Solving with Large Language Models17 May

Other references · 40

  1. 2017Attention Is All You Needcited by 59,932
  2. 2020Language Models are Few-Shot Learnerscited by 32,199
  3. 2023GPT-4 Technical Reportcited by 18,132
  4. 2022Training language models to follow instructions with human feedbackcited by 15,712
  5. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,865
  6. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,302
  7. 2021Training Verifiers to Solve Math Word Problemscited by 9,743
  8. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,601
  9. 2022PaLM: Scaling Language Modeling with Pathwayscited by 5,304
  10. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,203
  11. 2022Large Language Models are Zero-Shot Reasonerscited by 4,323
  12. 2021Finetuned Language Models Are Zero-Shot Learnerscited by 3,747
28 more
  1. 2022Scaling Instruction-Finetuned Language Modelscited by 2,997
  2. 2022Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language modelscited by 1,970
  3. 2022Challenging BIG-Bench Tasks and Whether Chain-of-Thought Can Solve Themcited by 1,767
  4. 2022Least-to-Most Prompting Enables Complex Reasoning in Large Language Modelscited by 1,397
  5. 2023PaLM 2 Technical Reportcited by 1,154
  6. 2022Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Taskscited by 1,057
  7. 2022Large Language Models Are Human-Level Prompt Engineerscited by 1,020
  8. 2023Graph of Thoughts: Solving Elaborate Problems with Large Language Modelscited by 873
  9. 2021Show Your Work: Scratchpads for Intermediate Computation with Language Modelscited by 871
  10. 2023Reasoning with Language Model is Planning with World Modelcited by 846
  11. 2023Large Language Models as Optimizerscited by 830
  12. 2021Cross-Task Generalization via Natural Language Crowdsourcing Instructionscited by 696
  13. 2022Decomposed Prompting: A Modular Approach for Solving Complex Taskscited by 554
  14. 2023Plan-and-Solve Prompting: Improving Zero-Shot Chain-of-Thought Reasoning by Large Language Modelscited by 550
  15. 2023Promptbreeder: Self-Referential Self-Improvement Via Prompt Evolutioncited by 521
  16. 2021Reframing Instructional Prompts to GPTk's Languagecited by 176
  17. 2023Take a Step Back: Evoking Reasoning via Abstraction in Large Language Modelscited by 157
  18. 2021Adapting Language Models for Zero-shot Learning by Meta-tuning on Dataset and Prompt Collectionscited by 128
  19. 2023Large Language Models as Analogical Reasonerscited by 107
  20. 2023Branch-Solve-Merge Improves Large Language Model Evaluation and Generationcited by 104
  21. 2022Compositional Semantic Parsing with Large Language Modelscited by 95
  22. 2023How FaR Are Large Language Models From Agents with Theory-of-Mind?cited by 75
  23. 2022Is a Question Decomposition Unit All We Need?cited by 37
  24. 2023Skills-in-Context Prompting: Unlocking Compositionality in Large Language Modelscited by 33
  25. 2023Plan, Verify and Switch: Integrated Reasoning with Diverse X-of-Thoughtscited by 27
  26. 2022Less is More: Summary of Long Instructions is Better for Program Synthesiscited by 17
  27. 2022HELP ME THINK: A Simple Prompting Strategy for Non-experts to Create Customized Content with Modelscited by 17
  28. 2023StrategyLLM: Large Language Models as Strategy Generators, Executors, Optimizers, and Evaluators for Problem Solvingcited by 13

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 33; 2025: 39; 2026: 36.

108 papers.

Sort:

By area:

  1. 2024From Generation to Judgment: Opportunities and Challenges of LLM-as-a-judgecited by 480
  2. 2024Automated Design of Agentic Systemscited by 360
  3. 2024Mutual Reasoning Makes Smaller LLMs Stronger Problem-Solverscited by 137
  4. 2025Agentic Large Language Models, a surveycited by 105
  5. 2024Buffer of Thoughts: Thought-Augmented Reasoning with Large Language Modelscited by 91
  6. 2024KnowAgent: Knowledge-Augmented Planning for LLM-Based Agentscited by 61
  7. 2025FlowReasoner: Reinforcing Query-Level Meta-Agentscited by 56
  8. 2025Atom of Thoughts for Markov LLM Test-Time Scalingcited by 54
  9. 2025A Survey of Large Language Model Agents for Question Answeringcited by 46
  10. 2024Agents Thinking Fast and Slow: A Talker-Reasoner Architecturecited by 35
  11. 2025Hogwild! Inference: Parallel LLM Generation via Concurrent Attentioncited by 32
  12. 2024LiteSearch: Efficacious Tree Search for LLMcited by 30
38 more
  1. 2024Are LLMs Effective Negotiators? Systematic Evaluation of the Multifaceted Capabilities of LLMs in Negotiation Dialoguescited by 29
  2. 2025MAS-ZERO: Designing Multi-Agent Systems with Zero Supervisioncited by 26
  3. 2024Instance-adaptive Zero-shot Chain-of-Thought Promptingcited by 25
  4. 2024Can LLMs Solve longer Math Word Problems Better?cited by 22
  5. 2025Position: Agent Should Invoke External Tools ONLY When Epistemically Necessarycited by 22
  6. 2024Language Models as Compilers: Simulating Pseudocode Execution Improves Algorithmic Reasoning in Language Modelscited by 20
  7. 2024RAD-Bench: Evaluating Large Language Models Capabilities in Retrieval Augmented Dialoguescited by 19
  8. 2024Step-by-Step Reasoning to Solve Grid Puzzles: Where do LLMs Falter?cited by 18
  9. 2024ReGenesis: LLMs can Grow into Reasoning Generalists via Self-Improvementcited by 17
  10. 2024DOTS: Learning to Reason Dynamically in LLMs via Optimal Reasoning Trajectories Searchcited by 17
  11. 2026A Comprehensive Survey on Agent Skills: Taxonomy, Techniques, and Applicationscited by 17
  12. 2024EPIC: Effective Prompting for Imbalanced-Class Data Synthesis in Tabular Data Classification via Large Language Modelscited by 14
  13. 2024HDFlow: Enhancing LLM Complex Problem-Solving with Hybrid Thinking and Dynamic Workflowscited by 13
  14. 2024P-FOLIO: Evaluating and Improving Logical Reasoning with Abundant Human-Written Reasoning Chainscited by 13
  15. 2025MetaScale: Test-Time Scaling with Evolving Meta-Thoughtscited by 13
  16. 2024Evaluating LLM Reasoning in the Operations Research Domain with ORQAcited by 12
  17. 2024Can LLM-Augmented autonomous agents cooperate?, An evaluation of their cooperative capabilities through Melting Potcited by 11
  18. 2024Enhancing Q&A with Domain-Specific Fine-Tuning and Iterative Reasoning: A Comparative Studycited by 11
  19. 2024HYBRIDMIND: Meta Selection of Natural Language and Symbolic Language for Enhanced LLM Reasoningcited by 11
  20. 2024Wrong-of-Thought: An Integrated Reasoning Framework with Multi-Perspective Verification and Wrong Informationcited by 10
  21. 2025Zero-Shot Verification-guided Chain of Thoughtscited by 9
  22. 2025Investigating the Shortcomings of LLMs in Step-by-Step Legal Reasoningcited by 9
  23. 2025LLMs for Customized Marketing Content Generation and Evaluation at Scalecited by 9
  24. 2024Cooperative Resilience in Artificial Intelligence Multiagent Systemscited by 8
  25. 2024PTD-SQL: Partitioning and Targeted Drilling with LLMs in Text-to-SQLcited by 7
  26. 2025Mitigating Hallucinations in Multimodal Spatial Relations through Constraint-Aware Promptingcited by 7
  27. 2025Long Context Scaling: Divide and Conquer via Multi-Agent Question-driven Collaborationcited by 7
  28. 2024IryoNLP at MEDIQA-CORR 2024: Tackling the Medical Error Detection & Correction Task On the Shoulders of Medical Agentscited by 6
  29. 2025Self-Reasoning Language Models: Unfold Hidden Reasoning Chains with Few Reasoning Catalystcited by 6
  30. 2025What Makes a Good Natural Language Prompt?cited by 6
  31. 2025Advanced Financial Reasoning at Scale: A Comprehensive Evaluation of Large Language Models on CFA Level IIIcited by 6
  32. 2026Learning with Rare Success but Rich Feedback via Reflection-Enhanced Self-Distillationcited by 6
  33. 2025DTS: Enhancing Large Reasoning Models via Decoding Tree Sketchingcited by 5
  34. 2026Stop Hand-Holding Your Coding Agent: Engineering the Loops that Replace Step-by-Step Promptingcited by 5
  35. 2024Auto-Evolve: Enhancing Large Language Model's Performance via Self-Reasoning Frameworkcited by 4
  36. 2025EditLord: Learning Code Transformation Rules for Code Editingcited by 4
  37. 2025Code-enabled language models can outperform reasoning models on diverse taskscited by 4
  38. 2025VICoT-Agent: A Vision-Interleaved Chain-of-Thought Framework for Interpretable Multimodal Reasoning and Scalable Remote Sensing Analysiscited by 4

50 most cited of 108 All 108 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2023Self-Polish: Enhance Reasoning in Large Language Models via Problem Refinement
  2. 2026Can We Break LLMs Out of Self-Loops? Fine-Grained Reasoning Control with Activation Steering
  3. 2025Advancing Reasoning in Large Language Models: Promising Methods and Approaches
  4. 2024Lachesis: Predicting LLM Inference Accuracy using Structural Properties of Reasoning Paths
  5. 2026A Decomposition Perspective to Long-context Reasoning for LLMs
  6. 2024Self-Explore: Enhancing Mathematical Reasoning in Language Models with Fine-grained Rewards
  7. 2025Lost at the Beginning of Reasoning
  8. 2024Toward Self-Improvement of LLMs via Imagination, Searching, and Criticizing