Understanding Layer Significance in LLM Alignment

Aligning large language models (LLMs) through fine-tuning is essential for tailoring them to specific applications. Therefore, understanding what LLMs learn during the alignment process is crucial. Recent studies suggest that alignment primarily adjusts a model's presentation style rather than its foundational knowledge, indicating that only certain components of the model are significantly impacted. To delve deeper into LLM alignment, we propose to identify which layers within LLMs are most critical to the alignment process, thereby uncovering how alignment influences model behavior at a granular level. We propose a novel approach to identify the important layers for LLM alignment (ILA). It involves learning a binary mask for each incremental weight matrix in the LoRA algorithm, indicating the significance of each layer. ILA consistently identifies important layers across various alignment datasets, with nearly 90% overlap even with substantial dataset differences, highlighting fundamental patterns in LLM alignment. Experimental results indicate that freezing non-essential layers improves overall model performance, while selectively tuning the most critical layers significantly enhances fine-tuning efficiency with minimal performance loss.

Parenting: Optimizing Knowledge Selection of Retrieval-Augmented Language Models with Parameter Decoupling and Tailored Tuning

Retrieval-Augmented Generation (RAG) offers an effective solution to the issues faced by Large Language Models (LLMs) in hallucination generation and knowledge obsolescence by incorporating externally retrieved knowledge. However, due to potential conflicts between internal and external knowledge, as well as retrieval noise, LLMs often struggle to effectively integrate external evidence, leading to a decline in performance. Although existing methods attempt to tackle these challenges, they often struggle to strike a balance between model adherence and robustness, resulting in significant learning variance. Inspired by human cognitive processes, we propose Parenting, a novel framework that decouples adherence and robustness within the parameter space of LLMs. Specifically, Parenting utilizes a key parameter mining method based on forward activation gain to identify and isolate the crucial parameter units that are strongly linked to adherence and robustness. Then, Parenting employs a type-guided tailored tuning strategy, applying specific and appropriate fine-tuning methods to parameter units representing different capabilities, aiming to achieve a balanced enhancement of adherence and robustness. Extensive experiments on various datasets and models validate the effectiveness and generalizability of our methods.

Diversity-grounded Channel Prototypical Learning for Out-of-Distribution Intent Detection

In the realm of task-oriented dialogue systems, a robust intent detection mechanism must effectively handle malformed utterances encountered in real-world scenarios. This study presents a novel fine-tuning framework for large language models (LLMs) aimed at enhancing in-distribution (ID) intent classification and out-of-distribution (OOD) intent detection, which utilizes semantic matching with prototypes derived from ID class names. By harnessing the highly distinguishable representations of LLMs, we construct semantic prototypes for each ID class using a diversity-grounded prompt tuning approach. We rigorously test our framework in a challenging OOD context, where ID and OOD classes are semantically close yet distinct, referred to as \emph{near} OOD detection. For a thorough assessment, we benchmark our method against the prevalent fine-tuning approaches. The experimental findings reveal that our method demonstrates superior performance in both few-shot ID intent classification and near-OOD intent detection tasks.

Continual Dialogue State Tracking via Reason-of-Select Distillation

An ideal dialogue system requires continuous skill acquisition and adaptation to new tasks while retaining prior knowledge. Dialogue State Tracking (DST), vital in these systems, often involves learning new services and confronting catastrophic forgetting, along with a critical capability loss termed the "Value Selection Quandary." To address these challenges, we introduce the Reason-of-Select (RoS) distillation method by enhancing smaller models with a novel 'meta-reasoning' capability. Meta-reasoning employs an enhanced multi-domain perspective, combining fragments of meta-knowledge from domain-specific dialogues during continual learning. This transcends traditional single-perspective reasoning. The domain bootstrapping process enhances the model's ability to dissect intricate dialogues from multiple possible values. Its domain-agnostic property aligns data distribution across different domains, effectively mitigating forgetting. Additionally, two novel improvements, "multi-value resolution" strategy and Semantic Contrastive Reasoning Selection method, significantly enhance RoS by generating DST-specific selection chains and mitigating hallucinations in teachers' reasoning, ensuring effective and reliable knowledge transfer. Extensive experiments validate the exceptional performance and robust generalization capabilities of our method. The source code is provided for reproducibility.

TaSL: Continual Dialog State Tracking via Task Skill Localization and Consolidation

A practical dialogue system requires the capacity for ongoing skill acquisition and adaptability to new tasks while preserving prior knowledge. However, current methods for Continual Dialogue State Tracking (DST), a crucial function of dialogue systems, struggle with the catastrophic forgetting issue and knowledge transfer between tasks. We present TaSL, a novel framework for task skill localization and consolidation that enables effective knowledge transfer without relying on memory replay. TaSL uses a novel group-wise technique to pinpoint task-specific and task-shared areas. Additionally, a fine-grained skill consolidation strategy protects task-specific knowledge from being forgotten while updating shared knowledge for bi-directional knowledge transfer. As a result, TaSL strikes a balance between preserving previous knowledge and excelling at new tasks. Comprehensive experiments on various backbones highlight the significant performance improvements of TaSL over existing state-of-the-art methods. The source code is provided for reproducibility.

TaSL: Task Skill Localization and Consolidation for Language Model Continual Learning

Language model continual learning (CL) has recently garnered significant interest due to its potential to adapt large language models (LLMs) to dynamic real-world environments without re-training. A key challenge in this field is catastrophic forgetting, where models lose previously acquired knowledge when learning new tasks. Existing methods commonly employ multiple parameter-efficient fine-tuning (PEFT) blocks to acquire task-specific knowledge for each task, but these approaches lack efficiency and overlook the potential for knowledge transfer through task interaction. In this paper, we present a novel CL framework for language models called Task Skill Localization and Consolidation (TaSL), which enhances knowledge transfer without relying on memory replay. TaSL first divides the model into `skill units' based on parameter dependencies, enabling more granular control. It then employs a novel group-wise skill localization technique to identify the importance distribution of skill units for a new task. By comparing this importance distribution with those from previous tasks, we implement a fine-grained skill consolidation strategy that retains task-specific knowledge, thereby preventing forgetting, and updates task-shared knowledge, which facilitates bi-directional knowledge transfer. As a result, TaSL achieves a superior balance between retaining previous knowledge and excelling in new tasks. TaSL also shows strong generalizability, suitable for general models and customizable for PEFT methods like LoRA. Additionally, it demonstrates notable extensibility, allowing integration with memory replay to further enhance performance. Extensive experiments on two CL benchmarks, with varying model sizes (from 220M to 7B), demonstrate the effectiveness of TaSL and its variants across different settings.

Towards LLM-driven Dialogue State Tracking

Dialogue State Tracking (DST) is of paramount importance in ensuring accurate tracking of user goals and system actions within task-oriented dialogue systems. The emergence of large language models (LLMs) such as GPT3 and ChatGPT has sparked considerable interest in assessing their efficacy across diverse applications. In this study, we conduct an initial examination of ChatGPT's capabilities in DST. Our evaluation uncovers the exceptional performance of ChatGPT in this task, offering valuable insights to researchers regarding its capabilities and providing useful directions for designing and enhancing dialogue systems. Despite its impressive performance, ChatGPT has significant limitations including its closed-source nature, request restrictions, raising data privacy concerns, and lacking local deployment capabilities. To address these concerns, we present LDST, an LLM-driven DST framework based on smaller, open-source foundation models. By utilizing a novel domain-slot instruction tuning method, LDST achieves performance on par with ChatGPT. Comprehensive evaluations across three distinct experimental settings, we find that LDST exhibits remarkable performance improvements in both zero-shot and few-shot setting compared to previous SOTA methods. The source code is provided for reproducibility.

A Multi-scale Generalized Shrinkage Threshold Network for Image Blind Deblurring in Remote Sensing

Remote sensing images are essential for many earth science applications, but their quality can be degraded due to limitations in sensor technology and complex imaging environments. To address this, various remote sensing image deblurring methods have been developed to restore sharp, high-quality images from degraded observational data. However, most traditional model-based deblurring methods usually require predefined hand-craft prior assumptions, which are difficult to handle in complex applications, and most deep learning-based deblurring methods are designed as a black box, lacking transparency and interpretability. In this work, we propose a novel blind deblurring learning framework based on alternating iterations of shrinkage thresholds, alternately updating blurring kernels and images, with the theoretical foundation of network design. Additionally, we propose a learnable blur kernel proximal mapping module to improve the blur kernel evaluation in the kernel domain. Then, we proposed a deep proximal mapping module in the image domain, which combines a generalized shrinkage threshold operator and a multi-scale prior feature extraction block. This module also introduces an attention mechanism to adaptively adjust the prior importance, thus avoiding the drawbacks of hand-crafted image prior terms. Thus, a novel multi-scale generalized shrinkage threshold network (MGSTNet) is designed to specifically focus on learning deep geometric prior features to enhance image restoration. Experiments demonstrate the superiority of our MGSTNet framework on remote sensing image datasets compared to existing deblurring methods.

How Good Are Large Language Models at Out-of-Distribution Detection?

Out-of-distribution (OOD) detection plays a vital role in enhancing the reliability of machine learning (ML) models. The emergence of large language models (LLMs) has catalyzed a paradigm shift within the ML community, showcasing their exceptional capabilities across diverse natural language processing tasks. While existing research has probed OOD detection with relative small-scale Transformers like BERT, RoBERTa and GPT-2, the stark differences in scales, pre-training objectives, and inference paradigms call into question the applicability of these findings to LLMs. This paper embarks on a pioneering empirical investigation of OOD detection in the domain of LLMs, focusing on LLaMA series ranging from 7B to 65B in size. We thoroughly evaluate commonly-used OOD detectors, scrutinizing their performance in both zero-grad and fine-tuning scenarios. Notably, we alter previous discriminative in-distribution fine-tuning into generative fine-tuning, aligning the pre-training objective of LLMs with downstream tasks. Our findings unveil that a simple cosine distance OOD detector demonstrates superior efficacy, outperforming other OOD detectors. We provide an intriguing explanation for this phenomenon by highlighting the isotropic nature of the embedding spaces of LLMs, which distinctly contrasts with the anisotropic property observed in smaller BERT family models. The new insight enhances our understanding of how LLMs detect OOD data, thereby enhancing their adaptability and reliability in dynamic environments.

Nest-DGIL: Nesterov-optimized Deep Geometric Incremental Learning for CS Image Reconstruction

Proximal gradient-based optimization is one of the most common strategies for solving image inverse problems as well as easy to implement. However, these techniques often generate heavy artifacts in image reconstruction. One of the most popular refinement methods is to fine-tune the regularization parameter to alleviate such artifacts, but it may not always be sufficient or applicable due to increased computational costs. In this work, we propose a deep geometric incremental learning framework based on second Nesterov proximal gradient optimization. The proposed end-to-end network not only has the powerful learning ability for high/low frequency image features,but also can theoretically guarantee that geometric texture details will be reconstructed from preliminary linear reconstruction.Furthermore, it can avoid the risk of intermediate reconstruction results falling outside the geometric decomposition domains and achieve fast convergence. Our reconstruction framework is decomposed into four modules including general linear reconstruction, cascade geometric incremental restoration, Nesterov acceleration and post-processing. In the image restoration step,a cascade geometric incremental learning module is designed to compensate for the missing texture information from different geometric spectral decomposition domains. Inspired by overlap-tile strategy, we also develop a post-processing module to remove the block-effect in patch-wise-based natural image reconstruction. All parameters in the proposed model are learnable,an adaptive initialization technique of physical-parameters is also employed to make model flexibility and ensure converging smoothly. We compare the reconstruction performance of the proposed method with existing state-of-the-art methods to demonstrate its superiority. Our source codes are available at https://github.com/fanxiaohong/Nest-DGIL.