A Lightweight Framework for Trigger-Guided LoRA-Based Self-Adaptation in LLMs

2025/09/05 by Jiacheng Wei, Wei, Jiacheng, Faguo Wu +3
Business, Management and Accounting · Computer Science · #Artificial Intelligence (cs.AI) #Business Process Modeling and Analysis #Computation and Language (cs.CL) #Context-Aware Activity Recognition Systems #FOS: Computer and information sciences

paper · pdf · doi:10.48550/arxiv.2509.05385

openalex publication_date 2025/09/05 · openalex created_date 2025/10/11 · openalex updated_date 2026/07/28

Abstract

Large language models are unable to continuously adapt and learn from new data during reasoning at inference time. To address this limitation, we propose that complex reasoning tasks be decomposed into atomic subtasks and introduce SAGE, a trigger-guided dynamic fine-tuning framework that enables adaptive updates during reasoning at inference time. SAGE consists of three key components: (1) a Trigger module that detects reasoning failures through multiple evaluation metrics in real time; (2) a Trigger Buffer module that clusters anomaly samples using a streaming clustering process with HDBSCAN, followed by stability checks and similarity-based merging; and (3) a Lora Store module that dynamically optimizes parameter updates with an adapter pool for knowledge retention. Evaluation results show that SAGE demonstrates excellent accuracy, robustness, and stability on the atomic reasoning subtask through dynamic knowledge updating during test time.

Citations

• How Much Can RAG Help the Reasoning of LLM?
• MLLM-CompBench: A Comparative Reasoning Benchmark for Multimodal LLMs
• Memory³: Language Modeling with Explicit Memory
• LoRA-Switch: Boosting the Efficiency of Dynamic LLM Adapters via System-Algorithm Co-design
• VB-LoRA: Extreme Parameter Efficient Fine-Tuning with Vector Banks
• Towards Modular LLMs by Building and Reusing a Library of LoRAs
• DyLoRA: Parameter Efficient Tuning of Pre-trained Models using Dynamic Search-Free Low-Rank Adaptation
• Training language models to follow instructions with human feedback
• Towards a Unified View of Parameter-Efficient Transfer Learning
• Transformer Feed-Forward Layers Are Key-Value Memories
• A Structured Self-attentive Sentence Embedding
• Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer
• Large-Margin Softmax Loss for Convolutional Neural Networks

Related

• Rank-1 LoRAs Encode Interpretable Reasoning Signals
• LoraMap: Harnessing the Power of LoRA Connections
• LoRA-PAR: A Flexible Dual-System LoRA Partitioning Approach to Efficient LLM Fine-Tuning
• Temporal Sampling for Forgotten Reasoning in LLMs
• DICE: Structured Reasoning in LLMs through SLM-Guided Chain-of-Thought Correction