mlsanityData to 7 Oct 2026

Implicit Reasoning in Transformers is Reasoning through Shortcuts

arXiv 2503.07604 ↗First posted 10 Mar 2025cs.CLcited by 14

On the map: LLM reasoning → Reasoning benchmarks and evaluation

Abstract (arXiv)

Test-time compute is emerging as a new paradigm for enhancing language models' complex multi-step reasoning capabilities, as demonstrated by the success of OpenAI's o1 and o3, as well as DeepSeek's R1. Compared to explicit reasoning in test-time compute, implicit reasoning is more inference-efficient, requiring fewer generated tokens. However, why does the advanced reasoning capability fail to emerge in the implicit reasoning style? In this work, we train GPT-2 from scratch on a curated multi-step mathematical reasoning dataset and conduct analytical experiments to investigate how language models perform implicit reasoning in multi-step tasks. Our findings reveal: 1) Language models can perform step-by-step reasoning and achieve high accuracy in both in-domain and out-of-domain tests via implicit reasoning. However, this capability only emerges when trained on fixed-pattern data. 2) Conversely, implicit reasoning abilities emerging from training on unfixed-pattern data tend to overfit a specific pattern and fail to generalize further. Notably, this limitation is also observed in state-of-the-art large language models. These findings suggest that language models acquire implicit reasoning through shortcut learning, enabling strong performance on tasks with similar patterns while lacking generalization.

Key facts. Cited by 14 papers in the map (data through 7 Oct 2026). Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning. Builds on 26 of its 32 references that are in the map. First posted on arXiv 10 Mar 2025.

Builds on

26 of 32 references.

References · 26

  1. 2021RoFormer: Enhanced Transformer with Rotary Position Embeddingcited by 4,581
  2. 2023SWE-bench: Can Language Models Resolve Real-World GitHub Issues?cited by 3,193
  3. 2023GPQA: A Graduate-Level Google-Proof Q&A Benchmarkcited by 2,748
  4. 2022Locating and Editing Factual Associations in GPTcited by 2,609
  5. 2022Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 smallcited by 1,130
  6. 2023Progress measures for grokking via mechanistic interpretabilitycited by 822
  7. 2023Finding Neurons in a Haystack: Case Studies with Sparse Probingcited by 328
  8. 2023Towards Best Practices of Activation Patching in Language Models: Metrics and Methodscited by 328
  9. 2023Does Localization Inform Editing? Surprising Differences in Causality-Based Localization vs. Knowledge Editing in Language Modelscited by 237
  10. 2024From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Stepcited by 225
  11. 2024Do Large Language Models Latently Perform Multi-Hop Reasoning?cited by 165
  12. 2023The Hydra Effect: Emergent Self-repair in Language Model Computationscited by 145
14 more
  1. 2023Physics of Language Models: Part 3.2, Knowledge Manipulationcited by 142
  2. 2023A Toy Model of Universality: Reverse Engineering How Networks Learn Group Operationscited by 141
  3. 2024Fine-Tuning Enhances Existing Mechanisms: A Case Study on Entity Trackingcited by 140
  4. 2024Physics of Language Models: Part 2.1, Grade-School Math and the Hidden Reasoning Processcited by 107
  5. 2024Universal Neurons in GPT2 Language Modelscited by 100
  6. 2024Arithmetic Without Algorithms: Language Models Solve Math With a Bag of Heuristicscited by 94
  7. 2024Hopping Too Late: Exploring the Limitations of Large Language Models on Multi-Hop Queriescited by 78
  8. 2022Measuring Causal Effects of Data Statistics on Language Model's `Factual' Predictionscited by 57
  9. 2024Interpreting and Improving Large Language Models in Arithmetic Calculationcited by 43
  10. 2024Pre-trained Large Language Models Use Fourier Features to Compute Additioncited by 38
  11. 2024Interpreting Arithmetic Mechanism in Large Language Models through Comparative Neuron Analysiscited by 31
  12. 2024Do Large Language Models Perform Latent Multi-Hop Reasoning without Exploiting Shortcuts?cited by 26
  13. 2024Revealing the Barriers of Language Agents in Planningcited by 20
  14. 2024Do LLMs Really Think Step-by-step In Implicit Reasoning?cited by 16

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 7; 2026: 7.

14 papers.

Sort:

By area:

  1. 2025A Survey of Efficient Reasoning for Large Reasoning Models: Language, Multimodality, and Beyondcited by 131
  2. 2025Advances and Challenges in Foundation Agents: From Brain-Inspired Intelligence to Evolutionary, Collaborative, and Safe Systemscited by 125
  3. 2025Reasoning Beyond Language: A Comprehensive Survey on Latent Chain-of-Thought Reasoningcited by 72
  4. 2025Implicit Reasoning in Large Language Models: A Comprehensive Surveycited by 42
  5. 2025Energy-Based Transformers are Scalable Learners and Thinkerscited by 30
  6. 2025Language Models, Graph Searching, and Supervision Adulteration: When More Supervision is Less and How to Make More Morecited by 3
  7. 2026Internalizing LLM Reasoning via Discovery and Replay of Latent Actionscited by 3
  8. 2026Neuron-Anchored Rule Extraction for Large Language Models via Contrastive Hierarchical Ablationcited by 2
  9. 2026BFS-PO: Best-First Search for Large Reasoning Modelscited by 1
  10. 2026Learning through Internalizationcited by 1
  11. 2026Final Checkpoints Are Not Enough: Analyzing Latent Reasoning Faithfulness Along Training Trajectoriescited by 1
  12. 2025The Kinetics of Reasoning: How Chain-of-Thought Shapes Learning in Transformers?cited by 0
2 more
  1. 2026Can We Trust LLM's Logic? Quantifying Uncertainty, Coherence, and Robustness via a Graph-Based Frameworkcited by 0
  2. 2026Detecting Hidden Chain-of-Thought in Large Language Models with Linguistic, Behavioral, and Mechanistic Indicatorscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025How do Transformers Learn Implicit Reasoning?
  2. 2024Grokked Transformers are Implicit Reasoners: A Mechanistic Journey to the Edge of Generalization
  3. 2026Loop, Think, & Generalize: Implicit Reasoning in Recurrent-Depth Transformers
  4. 2025Language models can learn implicit multi-hop reasoning, but only if they have lots of training data
  5. 2025iCLP: Large Language Model Reasoning with Implicit Cognition Latent Planning
  6. 2026Opening the Black Box: A Survey on the Mechanisms of Multi-Step Reasoning in Large Language Models
  7. 2023Implicit Chain of Thought Reasoning via Knowledge Distillation
  8. 2025Finding the Minimal Parameter Budget for Implicit Reasoning: A Data Complexity Driven Scaling Law for Language Models