Knowledge Tree
fine-tuning
Also known as: finetuning
Facts (54)
Sources
A Survey on the Theory and Mechanism of Large Language Models arxiv.org Mar 12, 2026 18 facts
claimThe training stage of an LLM pipeline consists of two processes: pre-training, which forges foundational capabilities, and fine-tuning, which adapts the model.
claimSparse Matrix Tuning is a development based on the insight that fine-tuning attention layers is more critical than tuning MLP layers, and it targets high-impact parameters.
claimLi et al. (2023c) validate that structural learning during fine-tuning is a robust phenomenon that persists even when training is restricted to specific components, affecting semantic organization.
claimTheoretical analysis challenges the assumption that fine-tuning fundamentally alters a large language model's knowledge or capabilities by investigating how alignment modifies the underlying model.
claimReinforcement learning on incorrect responses helps models identify and unlearn 'spurious correlations'—incorrect intermediate steps that lead to correct final answers—scaling synthetic dataset efficiency by eight-fold compared to standard positive-only fine-tuning.
measurementSetlur et al. (2024) found that in mathematical reasoning tasks, using reinforcement learning on a model's incorrect responses is twice as sample-efficient as fine-tuning on correct synthetic answers.
referenceThe paper 'Beyond zero initialization: investigating the impact of non-zero initialization on LoRA fine-tuning dynamics' was published in the Proceedings of the 42nd International Conference on Machine Learning, Vol. 267, pp. 35519–35535.
claimOuyang et al. (2022) and Ren and Sutherland (2025) identify fine-tuning as the critical process for adapting foundational models to specific tasks or human intent.
procedureThe DISCO framework, introduced by Zhang et al. (2025c), uses Singular Value Decomposition (SVD) to analyze a model’s features to avoid costly fine-tuning, based on the insight that different spectral components of features possess different degrees of transferability.
referenceThe paper 'SMT: fine-tuning large language models with sparse matrices' (The Thirteenth International Conference on Learning Representations) is cited in the survey 'A Survey on the Theory and Mechanism of Large Language Models' regarding fine-tuning.
claimThe research paper 'All roads lead to likelihood: the value of reinforcement learning in fine-tuning' (arXiv:2503.01067) analyzes the role and value of reinforcement learning in the fine-tuning process of large language models.
referenceZhu et al. (2024a) provided a systematic benchmark for assessing privacy leakage risks, which are exacerbated during model adaptation and fine-tuning.
claimYao et al. (2025c) provide a unified framework for selecting appropriate weight types and learning rates, offering theoretical guidance for the general fine-tuning of attention-based models.
procedureChoi et al. (2025) proposed the KDS framework, which quantifies contamination by measuring the change in the similarity structure of sample embeddings in the model’s representation space before and after fine-tuning, rather than using simple text matching.
claimHe et al. (2025a) provide empirical and theoretical evidence that fine-tuning attention layers is more critical for downstream tasks than tuning MLP layers.
claimBini et al. (2024) utilize orthogonal transformations for fine-tuning as an alternative to LoRA.
claimYuan et al. (2024) established the theoretical connection between LoRA and methods using orthogonal transformations for fine-tuning.
referenceThe paper 'The impact of initialization on lora finetuning dynamics' (Advances in Neural Information Processing Systems 37) is cited in the survey 'A Survey on the Theory and Mechanism of Large Language Models' regarding LoRA fine-tuning.
Unlocking the Potential of Generative AI through Neuro-Symbolic ... arxiv.org Feb 16, 2025 4 facts
claimIntegrating symbolic knowledge into neural network loss functions reinforces the connection between neural learning and symbolic reasoning in the contexts of model distillation, fine-tuning, pre-training, and transfer learning.
claimTransfer learning, which includes pre-training, fine-tuning, and few-shot learning, allows AI models to efficiently adapt knowledge from one task to another.
claimFeature extraction repurposes pre-trained model layers for specific tasks, while fine-tuning adjusts the weights of a pre-trained model for new tasks.
claimApproaches such as model distillation, fine-tuning, pre-training, and transfer learning align with the neuro-symbolic compiled paradigm by integrating symbolic constraints into the neural network learning process.
Medical Hallucination in Foundation Models and Their ... medrxiv.org Mar 3, 2025 4 facts
claimKnowledge editing techniques refine Large Language Model (LLM) outputs by directly modifying model weights or adding new knowledge parameters, rather than using iterative fine-tuning.
claimFine-tuning large language models on biomedical corpora significantly improves their understanding of clinical text, as demonstrated by Alsentzer et al. (2019).
claimTo ensure clinical relevance, Large Language Models require regular fine-tuning on updated medical data and integration with dynamic knowledge retrieval systems, such as tools capable of real-time evidence synthesis.
claimRobust finetuning procedures and retrieval-augmented generation can improve the balance of training data, which helps mitigate availability bias in large language models.
Awesome-Hallucination-Detection-and-Mitigation - GitHub github.com 4 facts
referenceThe paper 'Unfamiliar Finetuning Examples Control How Language Models Hallucinate' by Kang et al. (2024) investigates the impact of finetuning examples on hallucination behavior.
referenceThe paper 'Does Fine-Tuning LLMs on New Knowledge Encourage Hallucinations?' by Gekhman et al. (2024) examines the relationship between fine-tuning on new knowledge and hallucination rates.
referenceThe paper "Fine-tuning Language Models for Factuality" by Tian et al. (2023) discusses fine-tuning strategies specifically aimed at improving the factuality of language models.
referenceThe paper "Unfamiliar finetuning examples control how language" by Kang et al. (2024) examines how the use of unfamiliar finetuning examples influences the behavior of language models.
The construction and refined extraction techniques of knowledge ... nature.com Feb 10, 2026 3 facts
claimFull-parameter updates during fine-tuning can degrade the general knowledge previously acquired by a Large Language Model.
claimThe framework aims to balance lightweight fine-tuning of large language models (LLMs) with multi-task adaptability.
claimLarge-scale pre-trained Large Language Models (LLMs) such as GPT-4 and LLaMA-3 utilize large-scale pretraining and task-specific fine-tuning to achieve cross-task generalization.
The Synergy of Symbolic and Connectionist AI in LLM-Empowered ... arxiv.org Jul 11, 2024 3 facts
claimFine-tuning adapts pre-trained models to specific tasks or domains by utilizing a smaller, task-specific dataset to optimize performance for particular applications.
claimInstruction tuning and reinforcement learning from human feedback (RLHF) are proposed methods applied on top of fine-tuning to ensure Large Language Models follow human instructions, align with human values, and exhibit desired behaviors.
procedureThe training process for Large Language Models (LLMs) generally consists of two stages: pre-training and fine-tuning.
Survey and analysis of hallucinations in large language models frontiersin.org Sep 29, 2025 3 facts
claimGrounded pretraining and fine-tuning improves factual consistency by integrating knowledge sources or fact-labeled datasets during pretraining or fine-tuning stages, as noted by Zhang et al. (2023).
referenceLi et al. (2022) proposed fine-tuning methods that incorporate retrieved factual context to reduce hallucinations.
perspectiveFor developers deploying Large Language Models, selecting models based on attribution patterns (Prompt Sensitivity vs. Model Vulnerability) can inform fine-tuning strategies.
KG-RAG: Bridging the Gap Between Knowledge and Creativity - arXiv arxiv.org May 20, 2024 2 facts
claimRetrieval-Augmented Generation (RAG) can alleviate hallucinations and outperforms traditional fine-tuning methods for applications requiring high accuracy and up-to-date information by integrating external knowledge more effectively.
claimKnowledge Graphs enable Language Model Agents to access vast volumes of accurate and updated information without requiring resource-intensive fine-tuning.
On Hallucinations in Artificial Intelligence–Generated Content ... jnm.snmjournals.org 2 facts
claimTransfer learning, which involves leveraging publicly pretrained models and fine-tuning them on local data, is an effective strategy for balancing generalization and specialization to mitigate hallucinations.
claimContinuous dataset updating, which involves regularly fine-tuning models with newly acquired data to adapt to evolving clinical scenarios, introduces additional training costs and the risk of catastrophic forgetting.
A survey on augmenting knowledge graphs (KGs) with large ... link.springer.com Nov 4, 2024 2 facts
claimFine-tuning an LLM on embedded graph data aligns the model's general language understanding with the structured knowledge from the KG, which improves contextual features, increases reasoning capabilities, and reduces hallucinations.
claimFine-tuning large language models (LLMs) with knowledge graphs involves adapting pre-trained LLMs to use structured information from KGs to generate contextually accurate responses.
A Comprehensive Benchmark and Evaluation Framework for Multi ... arxiv.org Jan 6, 2026 1 fact
referenceA comparative analysis of medical AI implementation methods indicates that Prompt Engineering has very low implementation cost but low consistency, RAG has moderate implementation cost and high consistency, Fine-Tuning has high implementation cost and moderate consistency, and Multi-Agent systems have very high implementation cost and very high consistency.
Do LLMs Build World Representations? Probing Through ... neurips.cc Dec 9, 2024 1 fact
claimFine-tuning and advanced pre-training strengthen the tendency of large language models to maintain goal-oriented abstractions during decoding, which prioritizes task completion over the recovery of the world's state and dynamics.
LLM-KG4QA: Large Language Models and Knowledge Graphs for ... github.com 1 fact
referenceResearch on integrating Large Language Models with Knowledge Graphs is categorized into several distinct approaches: Pre-training, Fine-Tuning, KG-Augmented Prompting, Retrieval-Augmented Generation (RAG), Graph RAG, KG RAG, Hybrid RAG, Spatial RAG, Offline/Online KG Guidelines, Agent-based KG Guidelines, KG-Driven Filtering and Validation, Visual Question Answering (VQA), Multi-Document QA, Multi-Hop QA, Conversational QA, Temporal QA, Multilingual QA, Index-based Optimization, and Natural Language to Graph Query Language (NL2GQL).
Medical Hallucination in Foundation Models and Their Impact on ... medrxiv.org Nov 2, 2025 1 fact
perspectiveHallucination resistance in specialized medical contexts emerges from sophisticated reasoning capabilities, internal consistency mechanisms, and broad world knowledge developed during large-scale pretraining, rather than from domain-specific fine-tuning.
The Role of Hallucinations in Large Language Models - CloudThat cloudthat.com Sep 1, 2025 1 fact
procedureFine-tuning models with guardrails involves training the model to refuse to answer when uncertain and including rejection examples in the dataset, such as 'I’m not sure about that. Let me check the source before answering.'
Combining Knowledge Graphs and Large Language Models - arXiv arxiv.org Jul 9, 2024 1 fact
claimFuture research into combining knowledge graphs and large language models may address ineffective knowledge integration by modifying model architecture, fine-tuning, or injecting knowledge into feature-based pre-training models.
LLM Observability: How to Monitor AI When It Thinks in Tokens | TTMS ttms.com Feb 10, 2026 1 fact
claimMonitoring an AI system allows developers to identify categories of questions where the AI falters, enabling improvements such as fine-tuning or adding fallbacks, which increases user confidence and trust over time.
Unlock the Power of Knowledge Graphs and LLMs - TopQuadrant topquadrant.com 1 fact
claimKnowledge graphs improve the accuracy and contextual understanding of large language models and generative AI through retrieval-augmented generation (RAG), prompt-to-query techniques, or fine-tuning.
Unknown source 1 fact
claimThe KG-RAG framework is distinct from the process of fine-tuning a Large Language Model (LLM).