# Towards Understanding Fine-Tuning Mechanisms of LLMs via Circuit Analysis

arXiv: https://arxiv.org/abs/2502.11812 · HTML: https://mlsanity.com/paper/2502.11812/

## Key facts

- Cited by 20 papers in the map (data through 14 Oct 2026).
- Builds on 41 of its 51 references that are in the map.
- First posted on arXiv 17 Feb 2025.

## Builds on (most cited first, 30 of 41)

- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Scaling Instruction-Finetuned Language Models](https://mlsanity.com/paper/2210.11416/) (2022)
- [Locating and Editing Factual Associations in GPT](https://mlsanity.com/paper/2202.05262/) (2022)
- [Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling](https://mlsanity.com/paper/2304.01373/) (2023)
- [Transformer Feed-Forward Layers Are Key-Value Memories](https://mlsanity.com/paper/2012.14913/) (2020)
- [Inference-Time Intervention: Eliciting Truthful Answers from a Language Model](https://mlsanity.com/paper/2306.03341/) (2023)
- [Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small](https://mlsanity.com/paper/2211.00593/) (2022)
- [Mass-Editing Memory in a Transformer](https://mlsanity.com/paper/2210.07229/) (2022)
- [In-context Learning and Induction Heads](https://mlsanity.com/paper/2209.11895/) (2022)
- [Few-Shot Parameter-Efficient Fine-Tuning is Better and Cheaper than In-Context Learning](https://mlsanity.com/paper/2205.05638/) (2022)
- [Discovering Latent Knowledge in Language Models Without Supervision](https://mlsanity.com/paper/2212.03827/) (2022)
- [Towards Automated Circuit Discovery for Mechanistic Interpretability](https://mlsanity.com/paper/2304.14997/) (2023)
- [Transformer Feed-Forward Layers Build Predictions by Promoting Concepts in the Vocabulary Space](https://mlsanity.com/paper/2203.14680/) (2022)
- [Dissecting Recall of Factual Associations in Auto-Regressive Language Models](https://mlsanity.com/paper/2304.14767/) (2023)
- [Causal Abstractions of Neural Networks](https://mlsanity.com/paper/2106.02997/) (2021)
- [AdaLoRA: Adaptive Budget Allocation for Parameter-Efficient Fine-Tuning](https://mlsanity.com/paper/2303.10512/) (2023)
- [Does Localization Inform Editing? Surprising Differences in Causality-Based Localization vs. Knowledge Editing in Language Models](https://mlsanity.com/paper/2301.04213/) (2023)
- [Evaluating the Ripple Effects of Knowledge Editing in Language Models](https://mlsanity.com/paper/2307.12976/) (2023)
- [Retrieval Head Mechanistically Explains Long-Context Factuality](https://mlsanity.com/paper/2404.15574/) (2024)
- [Attribution Patching Outperforms Automated Circuit Discovery](https://mlsanity.com/paper/2310.10348/) (2023)
- [How does GPT-2 compute greater-than?: Interpreting mathematical abilities in a pre-trained language model](https://mlsanity.com/paper/2305.00586/) (2023)
- [Causal Mediation Analysis for Interpreting Neural NLP: The Case of Gender Bias](https://mlsanity.com/paper/2004.12265/) (2020)
- [A Mechanistic Understanding of Alignment Algorithms: A Case Study on DPO and Toxicity](https://mlsanity.com/paper/2401.01967/) (2024)
- [Fine-Tuning Enhances Existing Mechanisms: A Case Study on Entity Tracking](https://mlsanity.com/paper/2402.14811/) (2024)
- [Have Faith in Faithfulness: Going Beyond Circuit Overlap When Finding Model Mechanisms](https://mlsanity.com/paper/2403.17806/) (2024)
- [A Practical Review of Mechanistic Interpretability for Transformer-Based Language Models](https://mlsanity.com/paper/2407.02646/) (2024)
- [Mechanistically analyzing the effects of fine-tuning on procedurally defined tasks](https://mlsanity.com/paper/2311.12786/) (2023)
- [TruthX: Alleviating Hallucinations by Editing Large Language Models in Truthful Space](https://mlsanity.com/paper/2402.17811/) (2024)
- [A Mechanistic Interpretation of Arithmetic Reasoning in Language Models using Causal Mediation Analysis](https://mlsanity.com/paper/2305.15054/) (2023)
- [A Primer on the Inner Workings of Transformer-based Language Models](https://mlsanity.com/paper/2405.00208/) (2024)

## Built on by (most cited first, 20 of 20)

- [Model Unlearning via Sparse Autoencoder Subspace Guided Projections](https://mlsanity.com/paper/2505.24428/) (2025)
- [Does higher interpretability imply better utility? A Pairwise Analysis on Sparse Autoencoders](https://mlsanity.com/paper/2510.03659/) (2025)
- [Planted in Pretraining, Swayed by Finetuning: A Case Study on the Origins of Cognitive Biases in LLMs](https://mlsanity.com/paper/2507.07186/) (2025)
- [Mixture of Length and Pruning Experts for Knowledge Graphs Reasoning](https://mlsanity.com/paper/2507.20498/) (2025)
- [SafeSeek: Universal Attribution of Safety Circuits in Language Models](https://mlsanity.com/paper/2603.23268/) (2026)
- [Probability Signature: Bridging Data Semantics and Embedding Structure in Language Models](https://mlsanity.com/paper/2509.20124/) (2025)
- [PATCH: Mitigating PII Leakage in Language Models with Privacy-Aware Targeted Circuit PatcHing](https://mlsanity.com/paper/2510.07452/) (2025)
- [Demystifying Variance in Circuit Discovery of LLMs](https://mlsanity.com/paper/2606.16920/) (2026)
- [What does a Bayes-filtered transformer believe? A predictive Monte Carlo approach](https://mlsanity.com/paper/2607.17060/) (2026)
- [Text Capability Loss in Vision-Language Adaptation: An Attention-Sink Diagnosis](https://mlsanity.com/paper/2609.00746/) (2026)
- [Capsule Lens: Locating and Tracking Concept Geometry in Model Representations](https://mlsanity.com/paper/2609.05575/) (2026)
- [Distilled Circuits: A Mechanistic Study of Internal Restructuring in Knowledge Distillation](https://mlsanity.com/paper/2505.10822/) (2025)
- [Beyond Transfer Accuracy: Mechanism-Guided Controlled Adaptation for Low-Resource Languages](https://mlsanity.com/paper/2601.08146/) (2026)
- [Mechanism Shift During Post-training from Autoregressive to Masked Diffusion Language Models](https://mlsanity.com/paper/2601.14758/) (2026)
- [Training Prompt Matters: State-Adaptive Optimization for Robust Fine-Tuning](https://mlsanity.com/paper/2606.01967/) (2026)
- [Inherited Circuits, Learned Semantics: How Fine-Tuning Creates Evasion Vulnerabilities Invisible to Standard Evaluation](https://mlsanity.com/paper/2606.27091/) (2026)
- [AEGIS: A Mechanism-Guided Defense against Visual Synonym Jailbreaks in Text-to-Image Models](https://mlsanity.com/paper/2607.06120/) (2026)
- [Beyond Static Interpretability: Anticipating Post-SFT Mechanisms from Pre-SFT Parameters for Better Tuning](https://mlsanity.com/paper/2608.24482/) (2026)
- [Decoupling Internal Representational Changes and Causal Importance in Fine-Tuned Large Language Models](https://mlsanity.com/paper/2609.21113/) (2026)
- [Rewired or Gated? How Instruction Tuning Shapes Knowledge-Conflict Circuits in LLMs](https://mlsanity.com/paper/2609.25602/) (2026)

## Abstract (arXiv)

Fine-tuning significantly improves the performance of Large Language Models (LLMs), yet its underlying mechanisms remain poorly understood. This paper aims to provide an in-depth interpretation of the fine-tuning process through circuit analysis, a popular tool in Mechanistic Interpretability (MI). Unlike previous studies (Prakash et al. 2024; Chhabra et al. 2024) that focus on tasks where pre-trained models already perform well, we develop a set of mathematical tasks where fine-tuning yields substantial performance gains, which are closer to the practical setting. In our experiments, we identify circuits at various checkpoints during fine-tuning and examine the interplay between circuit analysis, fine-tuning methods, and task complexities. First, we find that while circuits maintain high node similarity before and after fine-tuning, their edges undergo significant changes, in contrast to prior work that shows circuits only add some additional components after fine-tuning. Based on these observations, we develop a circuit-aware Low-Rank Adaptation (LoRA) method, which assigns ranks to layers based on edge changes in the circuits. Experimental results demonstrate that our circuit-based LoRA algorithm achieves an average performance improvement of 2.46% over standard LoRA with similar parameter sizes. Furthermore, we explore how combining circuits from subtasks can enhance fine-tuning in compositional tasks, providing new insights into the design of such tasks and deepening the understanding of circuit dynamics and fine-tuning mechanisms.