The core idea is to utilize gradient information generated during optimization to dynamically adjust the noise intensity in a layer-wise and real-time manner to provide a simple yet effective solution for optimization-state aware dynamic regularization.
Abstract
Fine-tuning pre-trained language models (PLMs) is crucial for achieving performance gains in downstream natural language processing tasks, but it is prone to overfitting in low-resource scenarios. Existing noise-based regularization methods typically rely on static parameter statistics or predefined heuristics to mitigate this issue. However, the noise magnitude and injection timing in these approaches depend on fixed hyperparameters or heuristic schedules, making them unable to adapt to the dynamic optimization state of the model during training. To address this limitation, this paper proposes a novel fine-tuning framework named gradient-guided noise injection (GNI). The core idea is to utilize gradient information generated during optimization to dynamically adjust the noise intensity in a layer-wise and real-time manner. Extensive experiments on multiple tasks from the GLUE and SuperGLUE benchmarks demonstrate that GNI generally improves average performance across different pre-trained models and task settings, while exhibiting stable generalizability and robustness. This work provides a simple yet effective solution for optimization-state aware dynamic regularization.
Fine-tuning large language models (LLMs) is an effective approach to enhancing their performance on specialized downstream tasks. Among the various techniques, low-rank adaptation has garnered significant attention due to its ability to maintain the full performance of fine-tuning while enhancing computational efficiency. However, existing approaches often rely on manually specified and fixed hyperparameters to identify the trainable components within weight matrices, resulting in suboptimal performance and low parameter efficiency. This paper presents a novel Le arnable Lo w-R ank A daptation (LeLoRA) framework that utilizes dynamically learned fine-tuning strategies to facilitate the effective adaptation of LLMs. Our framework integrates an LLM with a policy network that automatically and adaptively generates matrix-specific adaptation strategies to identify the trainable components of each weight matrix, taking into account their unique characteristics, such as singular values and matrix norms. A reinforcement learning-based optimization algorithm is then employed to iteratively update the LLM and the policy network, ensuring that the generated strategies adapt in real time to the evolving states of the LLM. Extensive experiments have been conducted across various natural language processing tasks. The results across ten different LLMs, ranging from 125M to 70B parameters, provide compelling evidence that LeLoRA consistently outperforms existing baselines in adapting LLMs.
Xiaoling Zhou, Mingjie Zhang, Zhemg Lee et al.· Annual Meeting of the Associ...· 0 citations
Pruning is essential for the efficient deployment of Large Language Models (LLMs); however, it causes severe performance degradation due to the structural distortion induced by sparsity. Existing recovery strategies, such as LoRA, predominantly employ global fine-tuning, often overlooking the mechanistic root of this degradation: the layer-wise accumulation and amplification of local errors. To address this limitation, we propose LaCo ( La yer-wise Co mpensation), a framework that reori-ents the recovery paradigm from global adaptation to hierarchical representation alignment. By sequentially optimizing each layer to re-construct the model’s hidden states, LaCo effectively intercepts the error propagation chain at its source. Extensive experiments demonstrate that LaCo surpasses parameter-efficient baselines in both perplexity reduction and zero-shot reasoning. Notably, it reduces recovery-time memory usage to approximately 1 / 7 of the baseline and requires only 2,048 unlabeled samples to match a LoRA model trained on 50k examples—achieving a ∼ 25 × improvement in data efficiency.
Yingen Liu, Fan Wu, Xuyan Pan et al.· Annual Meeting of the Associ...· 0 citations
SAMPLe (Sharpness-Aware Minimization Prompt Learning), a plug-in sharpness-aware optimizer that enhances prompt generalizability by accounting for loss landscape sharpness, is introduced, establishing itself as a robust, model-agnostic solution for prompt learning.
Fine-tuning large pre-trained models on specific tasks can improve in-distribution (ID) accuracy but often degrades out-of-distribution (OOD) generalization. Discrepancy-Guided Parameter Suppression (DPS) is a fine-tuning approach that Discrepancy-Guided Parameter Suppression (DPS) is a fine-tuning approach that freezes the top K percent of model parameters based on the magnitude of their changes. By ranking parameters according to a discrepancy score (the magnitude of change during fine-tuning) and suppressing the highest-discrepancy parameters, DPS encourages the model to learn task-relevant features without overfitting to spurious ID-specific patterns. We evaluate DPS on the challenging iWildCam dataset under camera-induced domain shifts, providing the first in-depth assessment of DPS on a real-world distribution shift benchmark. Using a CLIP ViT-B/16 backbone and identical hyperparameters for a cross-entropy baseline and DPS, we observe improvements on OOD metrics: OOD macro-F1 increases from 35.14% to 36.15% and OOD accuracy from 73.92% to 74.95%, roughly a 1% absolute gain on both. We also provide ablation analyses of the discrepancy metric and suppression rate, showing that normalized, channel-wise suppression at high freezing ratios yields the strongest OOD robustness, consistent with prior findings.
Chak Pang Hong· International Conference on...· 0 citations
Fine-tuning Large Language Models (LLMs) has become a crucial technique for adapting pre-trained models to downstream tasks. However, the enormous size of LLMs poses significant challenges in terms of computational complexity and resource requirements. Low-Rank Adaptation (LoRA) has emerged as a promising solution, yet a gap remains between the practical performance of low-rank adaptations and their theoretical optimum. While recent works have explored iteratively merging LoRA adapters, they lack a principled theoretical framework to guide adapter design. In this work, we propose eXtreme Gradient Boosting LoRA (XGBLoRA), a novel framework grounded in gradient boosting theory. Our key insight is that the adapter must adhere to the weak learner principle-each individual adapter should have limited expressiveness---to ensure that the iterative ensemble can effectively raise the model's performance ceiling without overfitting. We provide theoretical analysis establishing convergence guarantees and expressiveness bounds, which formally justify why weaker (lower-rank) adapters, when properly combined, can match or exceed the performance of stronger (higher-rank) ones. Extensive experiments on natural language processing tasks demonstrate that XGBLoRA with rank-1 updates consistently outperforms standard LoRA with significantly fewer trainable parameters.
Yifei Zhang, Hao Zhu, Haoran Shi et al.· Proceedings of the 32nd ACM...· 0 citations
A retraining-free VLM pruning framework called PORTA is introduced that derives a task- and modality-agnostic importance formulation based on activation variation, estimated from generic calibration data, which reliably captures feature-level representation utility across modalities.
Minseok Kang, Hyunwoo J. Kim, Chanyoung Kim et al.· 0 citations