mlsanityData to 14 Oct 2026

Competition of Mechanisms: Tracing How Language Models Handle Facts and Counterfactuals

arXiv 2402.11655 ↗First posted 18 Feb 2024cs.CLcited by 40

Not on the map. Why?

Abstract (arXiv)

Interpretability research aims to bridge the gap between empirical success and our scientific understanding of the inner workings of large language models (LLMs). However, most existing research focuses on analyzing a single mechanism, such as how models copy or recall factual knowledge. In this work, we propose a formulation of competition of mechanisms, which focuses on the interplay of multiple mechanisms instead of individual mechanisms and traces how one of them becomes dominant in the final prediction. We uncover how and where mechanisms compete within LLMs using two interpretability methods: logit inspection and attention modification. Our findings show traces of the mechanisms and their competition across various model components and reveal attention positions that effectively control the strength of certain mechanisms. Code: https://github.com/francescortu/comp-mech. Data: https://huggingface.co/datasets/francescortu/comp-mech.

Key facts. Cited by 40 papers in the map (data through 14 Oct 2026). Builds on 6 of its 15 references that are in the map. First posted on arXiv 18 Feb 2024.

Builds on

6 of 15 references.

References · 6

  1. 2019How Can We Know What Language Models Know?cited by 959
  2. 2023Progress measures for grokking via mechanistic interpretabilitycited by 839
  3. 2023How does GPT-2 compute greater-than?: Interpreting mathematical abilities in a pre-trained language modelcited by 202
  4. 2022Inverse scaling can become U-shapedcited by 76
  5. 2023Copy Suppression: Comprehensively Understanding an Attention Headcited by 76
  6. 2020Hierarchical nucleation in deep neural networkscited by 31

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2024: 12; 2025: 15; 2026: 13.

40 papers.

Sort:

By area:

  1. 2024A Primer on the Inner Workings of Transformer-based Language Modelscited by 85
  2. 2024Attention Heads of Large Language Models: A Surveycited by 67
  3. 2024Steering Knowledge Selection Behaviours in LLMs via SAE-Based Representation Engineeringcited by 63
  4. 2024Activation Scaling for Steering and Interpreting Language Modelscited by 35
  5. 2024Interpreting Key Mechanisms of Factual Recall in Transformer-Based Language Modelscited by 31
  6. 2024Intrinsic Test of Unlearning Using Parametric Knowledge Tracescited by 29
  7. 2024Controllable Context Sensitivity and the Knob Behind Itcited by 23
  8. 2024Analysing the Residual Stream of Language Models Under Knowledge Conflictscited by 17
  9. 2025Head Pursuit: Probing Attention Specialization in Multimodal Transformerscited by 17
  10. 2024Counterfactual Token Generation in Large Language Modelscited by 14
  11. 2025Understanding and Leveraging the Expert Specialization of Context Faithfulness in Mixture-of-Experts LLMscited by 12
  12. 2025The Atlas of In-Context Learning: How Attention Heads Shape In-Context Retrieval Augmentationcited by 11
28 more
  1. 2025When Seeing Overrides Knowing: Disentangling Knowledge Conflicts in Vision-Language Modelscited by 11
  2. 2025Repetitions are not all alike: distinct mechanisms sustain repetition in language modelscited by 10
  3. 2025Reasoning or Retrieval? A Study of Answer Attribution on Large Reasoning Modelscited by 10
  4. 2024Generalisation First, Memorisation Second? Memorisation Localisation for Natural Language Classification Taskscited by 9
  5. 2024A Reality Check on Context Utilisation for Retrieval-Augmented Generationcited by 9
  6. 2025Why Safeguarded Ships Run Aground? Aligned Large Language Models' Safety Mechanisms Tend to Be Anchored in The Template Regioncited by 6
  7. 2026Instruction Anchor: Dissecting the Mechanistic Dynamics of Modality Arbitrationcited by 6
  8. 2026Where Knowledge Collides: A Mechanistic Study of Intra-Memory Knowledge Conflict in Language Modelscited by 5
  9. 2025CUB: Benchmarking Context Utilisation Techniques for Language Modelscited by 4
  10. 2025Non-Interactive Symbolic-Aided Chain-of-Thought for Logical Reasoningcited by 4
  11. 2025Understanding Fact Recall in Language Models: Why Two-Stage Training Encourages Memorization but Mixed Training Teaches Knowledgecited by 1
  12. 2025Neuron-Level Analysis of Cultural Understanding in Large Language Modelscited by 1
  13. 2025Where does an LLM begin computing an instruction?cited by 1
  14. 2026To Copy or Not to Copy: Copying Is Easier to Induce Than Recallcited by 1
  15. 2026Diagnosing LLM Arbitration Behavior over Pre-evidence Epistemic States in RAG-based Fact-Checkingcited by 1
  16. 2024Identifying Sub-networks in Neural Networks via Functionally Similar Representationscited by 0
  17. 2025Time Course MechInterp: Analyzing the Evolution of Components and Knowledge in Large Language Modelscited by 0
  18. 2025On the Generalizability of "Competition of Mechanisms: Tracing How Language Models Handle Facts and Counterfactuals"cited by 0
  19. 2025Tracing Facts or just Copies? A critical investigation of the Competitions of Mechanisms in Large Language Modelscited by 0
  20. 2026TRUE: A Trustworthy Unified Explanation Framework for Large Language Model Reasoningcited by 0
  21. 2026Persistent Priors, Preserved Targets: A Stroop-Style Paradigm for Lexical Overridecited by 0
  22. 2026Attention-Path Fragility as an Uncertainty Signal in Large Language Modelscited by 0
  23. 2026How Do Language Models Choose Between Context and Memory?cited by 0
  24. 2026From Vision to Language: Investigating Causal Information Flow in Multimodal Decision-Makingcited by 0
  25. 2026Stale-Document Poisoning: When Outdated Retrieval Overrides Correct Model Answerscited by 0
  26. 2026Generalization Dynamics of LM Pre-trainingcited by 0
  27. 2026Don't Forget! Decomposing the Training Dynamics of Memorization in Language Modelscited by 0
  28. 2026Spontaneous Context Restoration: How Language Models Recover from Corrupted Inputscited by 0