Large Language Models (LLMs) have brought significant advances across various NLP tasks through few-shot or zero-shot prompting, bypassing the need for parameter tuning. However, the "black-box" nature behind their massive parameter sizes increases the "hallucination" concerns, especially in high-stakes applications (e.g., healthcare), where decision mistakes can lead to severe consequences. In contrast, human decision-making relies on complex cognitive processes, such as the ability to sense and adaptively correct mistakes through conceptual understanding. Drawing inspiration from human cognition, we propose an innovative metacognitive approach CLEAR, to equip LLMs with capabilities for self-aware error identification and correction. Our framework constructs concept-specific sparse subnetworks that indicate decision processes. This provides a novel interface for model {intervention} after deployment. The benefits include: (i) at inference time, our metacognitive LLMs can self-consciously identify potential mispredictions with minimum human involvement, (ii) the model can self-correct its errors efficiently without additional tuning, and (iii) the correction procedure is not only self-explanatory but also user-friendly, enhancing model interpretability and accessibility. With these metacognitive features, our approach pioneers a new path toward the trustworthiness of LLMs.
论文检索
输入标题、作者或关键词，从 7,537 篇学术成果中精准定位
Text-to-SQL enables users to interact with databases through natural language, simplifying the retrieval and synthesis of information. Despite the success of large language models (LLMs) in converting natural language questions into SQL queries, their broader adoption is limited by two main challenges: achieving robust generalization across diverse queries and ensuring interpretative confidence in their predictions. To tackle these issues, our research investigates the integration of selective classifiers into Text-to-SQL systems. We analyse the trade-off between coverage and risk using entropy based confidence estimation with selective classifiers and assess its impact on the overall performance of Text-to-SQL models. Additionally, we explore the models' initial calibration and improve it with calibration techniques for better model alignment between confidence and accuracy. Our experimental results show that encoder-decoder T5 is better calibrated than in-context-learning GPT 4 and decoder-only Llama 3, thus the designated external entropy-based selective classifier has better performance. The study also reveal that, in terms of error detection, selective classifier with a higher probability detects errors associated with irrelevant questions rather than incorrect query generations.
CLNX: Bridging Code and Natural Language for C/C++ Vulnerability-Contributing Commits Identification
PDF ↗Large Language Models (LLMs) have shown great promise in vulnerability identification. As C/C++ comprise half of the open-source Software (OSS) vulnerabilities over the past decade and updates in OSS mainly occur through commits, enhancing LLMs' ability to identify C/C++ Vulnerability-Contributing Commits (VCCs) is essential. However, current studies primarily focus on further pre-training LLMs on massive code datasets, which is resource-intensive and poses efficiency challenges. In this paper, we enhance the ability of BERT-based LLMs to identify C/C++ VCCs in a lightweight manner. We propose CodeLinguaNexus (CLNX) as a bridge facilitating communication between C/C++ programs and LLMs. Based on commits, CLNX efficiently converts the source code into a more natural representation while preserving key details. Specifically, CLNX first applies Structure-level Naturalization to decompose complex programs, followed by Token-level Naturalization to interpret complex symbols. We evaluate CLNX on public datasets of 25,872 C/C++ functions with their commits. The results demonstrate that CLNX substantially improves the ability of LLMs to detect C/C++ VCCs. Moreover, CLNX-equipped CodeBERT achieves new state-of-the-art performance and identifies 38 OSS vulnerabilities in the real world.
Divide-Solve-Combine: An Interpretable and Accurate Prompting Framework for Zero-shot Multi-Intent Detection
PDF ↗Zero-shot multi-intent detection is capable of capturing multiple intents within a single utterance without any training data, which gains increasing attention. Building on the success of large language models (LLM), dominant approaches in the literature explore prompting techniques to enable zero-shot multi-intent detection. While significant advancements have been witnessed, the existing prompting approaches still face two major issues: lacking explicit reasoning and lacking interpretability. Therefore, in this paper, we introduce a Divide-Solve-Combine Prompting (DSCP) to address the above issues. Specifically, DSCP explicitly decomposes multi-intent detection into three components including (1) single-intent division prompting is utilized to decompose an input query into distinct sub-sentences, each containing a single intent; (2) intent-by-intent solution prompting is applied to solve each sub-sentence recurrently; and (3) multi-intent combination prompting is employed for combining each sub-sentence result to obtain the final multi-intent result. By decomposition, DSCP allows the model to track the explicit reasoning process and improve the interpretability. In addition, we propose an interactive divide-solve-combine prompting (Inter-DSCP) to naturally capture the interaction capabilities of large language models. Experimental results on two standard multi-intent benchmarks (i.e., MixATIS and MixSNIPS) reveal that both DSCP and Inter-DSCP obtain substantial improvements over baselines, achieving superior performance and higher interpretability.
Significant efforts have been dedicated to integrating the powerful Large Language Models (LLMs) with diverse modalities, particularly focusing on the fusion of language, vision and audio data. However, the graph-structured data, which is inherently rich in structural and domain-specific knowledge, has not yet been gracefully adapted to LLMs. Existing methods either describe the graph with raw text, suffering the loss of graph structural information, or feed Graph Neural Network (GNN) embeddings into LLMs at the cost of losing explainable prompt semantics. To bridge this gap, we introduce an end-to-end modality-aligning framework for LLM-graph alignment: Dual-Residual Vector Quantized-Variational AutoEncoder, namely Dr.E. Our approach is purposefully designed to facilitate token-level alignment with LLMs, enabling an effective translation of the intrinsic `language' of graphs into comprehensible natural language. We also manage to enhance LLMs' more robust structural understanding of graphs by incorporating multiple views of the central nodes based on their surrounding nodes at various distances. Our experimental evaluations on standard graph tasks demonstrate competitive performance against other state-of-the-art (SOTA) approaches. Additionally, our framework ensures certain visual interpretability, efficiency, and robustness, marking the promising successful endeavor to achieve token-level alignment between LLMs and GNNs.
Relation Also Knows: Rethinking the Recall and Editing of Factual Associations in Auto-Regressive Transformer Language Models
PDF ↗The storage and recall of factual associations in auto-regressive transformer language models (LMs) have drawn a great deal of attention, inspiring knowledge editing by directly modifying the located model weights. Most editing works achieve knowledge editing under the guidance of existing interpretations of knowledge recall that mainly focus on subject knowledge. However, these interpretations are seriously flawed, neglecting relation information and leading to the *over-generalizing* problem for editing. In this work, we discover a novel relation-focused perspective to interpret the knowledge recall of transformer LMs during inference and apply it on single knowledge editing to avoid over-generalizing. Experimental results on the dataset supplemented with a new R-Specificity criterion demonstrate that our editing approach significantly alleviates over-generalizing while remaining competitive on other criteria, breaking the domination of subject-focused editing for future research.
KAES: Multi-aspect Shared Knowledge Finding and Aligning for Cross-prompt Automated Scoring of Essay Traits
PDF ↗Cross-prompt automated essay scoring (AES) aims to train models using essays from different source prompts and test them on new target prompt essays. A core challenge of the task is to learn as much shared knowledge as possible between essays from different prompts in order to better represent new prompt essays. Previous studies primarily focus on learning this knowledge on a general, coarse-grained level, ignoring that the shared knowledge among prompts is highly detailed and contains a more comprehensive range of information that is not fully investigated. In this paper, we propose a novel multi-aspect knowledge finding and aligning optimization strategy to better acquire this detailed various shared knowledge. We also introduce LLM to extract explicit, interpretable knowledge from implicit, multi-aspect shared knowledge and use this knowledge to improve the representation and evaluation performance of new prompt essays. We conduct extensive experiments on public datasets. The results show that our approach outperforms current state-of-the-art models and is effective on cross-prompt AES.
Model editing aims at selectively updating a small subset of a neural model's parameters with an interpretable strategy to achieve desired modifications. It can significantly reduce computational costs to adapt to large language models(LLMs). Given its ability to precisely target critical components within LLMs, model editing shows great potential for efficient fine-tuning applications. In this work, we investigate model editing to serve as an efficient method for adapting LLMs to solve aspect-based sentiment classification. Through causal interventions, we trace and determine which neuron hidden states are essential for the model’s predictions. By performing interventions and restorations on each component of an LLM, we identify the importance of these components for aspect-based sentiment classification. Our findings reveal that a distinct set of mid-layer representations is essential for detecting the sentiment polarity of given aspect words. Leveraging these insights, we develop a model editing approach that focuses exclusively on these critical parts of the LLM, leading to a more efficient method for adapting LLMs. Our in and out of domain experiments demonstrate that this approach achieves competitive results compared to the currently strongest methods with significantly fewer trainable parameters, highlighting a more efficient and interpretable fine-tuning strategy.
Generating Chain-of-Thought (CoT) before deriving the answer can effectively improve the reasoning capabilities of large language models (LLMs) and significantly improve the accuracy of the generated answer. However, in most cases, the length of the generated CoT is much longer than the desired final answer, which results in additional decoding costs. Furthermore, existing research has discovered that shortening the reasoning steps in CoT, even while preserving the key information, diminishes LLMs' abilities. These phenomena make it difficult to use LLMs and CoT in many real-world applications that only require the final answer and are sensitive to latency, such as search and recommendation. To reduce the costs of model decoding and shorten the length of the generated CoT, this paper presents Conditioned Compressed Chain-of-Thought (C3oT), a CoT compression framework that involves a compressor to compress an original longer CoT into a shorter CoT while maintaining key information and interpretability, a conditioned training method to train LLMs with both longer CoT and shorter CoT simultaneously to learn the corresponding relationships between them, and a conditioned inference method to gain the reasoning ability learned from longer CoT by generating shorter CoT. We conduct experiments over four datasets from arithmetic and commonsense scenarios, showing that the proposed method is capable of compressing the length of generated CoT by up to more than 50% without compromising its effectiveness.
Leveraging the Dual Capabilities of LLM: LLM-Enhanced Text Mapping Model for Personality Detection
PDF ↗Personality detection aims to deduce a user’s personality from their published posts. The goal of this task is to map posts to specific personality types. Existing methods encode post information to obtain user vectors, which are then mapped to personality labels. However, existing methods face two main issues: first, only using small models makes it hard to accurately extract semantic features from multiple long documents. Second, the relationship between user vectors and personality labels is not fully considered. To address the issue of poor user representation, we utilize the text embedding capabilities of LLM. To solve the problem of insufficient consideration of the relationship between user vectors and personality labels, we leverage the text generation capabilities of LLM. Therefore, we propose the LLM-Enhanced Text Mapping Model (ETM) for Personality Detection. The model applies LLM’s text embedding capability to enhance user vector representations. Additionally, it uses LLM’s text generation capability to create multi-perspective interpretations of the labels, which are then used within a contrastive learning framework to strengthen the mapping of these vectors to personality labels. Experimental results show that our model achieves state-of-the-art performance on benchmark datasets.
Self-correction in text-to-SQL is the process of prompting large language model (LLM) to revise its previously incorrectly generated SQL, and commonly relies on manually crafted self-correction guidelines by human experts that are not only labor-intensive to produce but also limited by the human ability in identifying all potential error patterns in LLM responses. We introduce MAGIC, a novel multi-agent method that automates the creation of the self-correction guideline. MAGIC uses three specialized agents: a manager, a correction, and a feedback agent. These agents collaborate on the failures of an LLM-based method on the training set to iteratively generate and refine a self-correction guideline tailored to LLM mistakes, mirroring human processes but without human involvement. Our extensive experiments show that MAGIC's guideline outperforms expert human's created ones. We empirically find out that the guideline produced by MAGIC enhances the interpretability of the corrections made, providing insights in analyzing the reason behind the failures and successes of LLMs in self-correction.
Emergent Communication (EC) provides a unique window into the language systems that emerge autonomously when agents are trained to jointly achieve shared goals. However, it is difficult to interpret EC and evaluate its relationship with natural languages (NL). This study employs unsupervised neural machine translation (UNMT) techniques to decipher ECs formed during referential games with varying task complexities, influenced by the semantic diversity of the environment. Our findings demonstrate UNMT's potential to translate EC, illustrating that task complexity characterized by semantic diversity enhances EC translatability, while higher task complexity with constrained semantic variability exhibits pragmatic EC, which, although challenging to interpret, remains suitable for translation. This research marks the first attempt, to our knowledge, to translate EC without the aid of parallel data.
Decision trees are widely used in machine learning due to their simplicity and interpretability, but they often lack robustness to adversarial attacks and data perturbations. The paper proposes a novel island-based coevolutionary algorithm (ICoEvoRDF) for constructing robust decision tree ensembles. The algorithm operates on multiple islands, each containing populations of decision trees and adversarial perturbations. The populations on each island evolve independently, with periodic migration of top-performing decision trees between islands. This approach fosters diversity and enhances the exploration of the solution space, leading to more robust and accurate decision tree ensembles. ICoEvoRDF utilizes a popular game theory concept of mixed Nash equilibrium for ensemble weighting, which further leads to improvement in results. ICoEvoRDF is evaluated on 20 benchmark datasets, demonstrating its superior performance compared to state-of-the-art methods in optimizing both adversarial accuracy and minimax regret. The flexibility of ICoEvoRDF allows for the integration of decision trees from various existing methods, providing a unified framework for combining diverse solutions. Our approach offers a promising direction for developing robust and interpretable machine learning models.
Time series forecasting plays a crucial role in domains such as finance, healthcare, and climate science. However, as modern time series data become increasingly complex, featuring high dimensionality, intricate spatiotemporal dependencies, and multi-scale evolutionary patterns, traditional analytical methods and existing predictive models face significant challenges. Although Large Language Models (LLMs) excel in capturing long-range dependencies, they still struggle with multi-scale dynamics and seasonal patterns. Moreover, while LLMs' semantic representation capabilities are rich, they often lack explicit alignment with the numerical patterns and temporal structures of time series data, leading to limitations in predictive accuracy and interpretability. To address these challenges, this paper proposes a novel framework, STEM-LTS (Semantic-TEmporal Modeling for Large-scale Time Series). STEM-LTS enhances the ability to capture complex spatiotemporal dependencies by integrating time series decomposition techniques with LLM-based modeling. The semantic-temporal alignment mechanism within the framework significantly improves LLMs' ability to interpret and forecast time series data. Additionally, we develop an adaptive multi-task learning strategy to optimize the model's performance across multiple dimensions. Through extensive experiments on various real-world datasets, we demonstrate that STEM-LTS achieves significant improvements in prediction accuracy, robustness to noise, and interpretability. Our work not only advances LLM-based time series analysis but also offers new perspectives on handling complex temporal data.
The Information Bottleneck (IB) principle has emerged as a promising approach for enhancing the generalization, robustness, and interpretability of deep neural networks, demonstrating efficacy across image segmentation, document clustering, and semantic communication. Among IB implementations, the IB Lagrangian method, employing Lagrangian multipliers, is widely adopted. While numerous methods for the optimizations of IB Lagrangian based on variational bounds and neural estimators are feasible, their performance is highly dependent on the quality of their design, which is inherently prone to errors. To address this limitation, we introduce Structured IB, a framework for investigating potential structured features. By incorporating auxiliary encoders to extract missing informative features, we generate more informative representations. Our experiments demonstrate superior prediction accuracy and task-relevant information preservation compared to the original IB Lagrangian method, even with reduced network size.
Recent concept-based interpretable models have succeeded in providing meaningful explanations by pre-defined concept sets. However, the dependency on the pre-defined concepts restricts the application because of the limited number of concepts for explanations. This paper proposes a novel interpretable deep neural network called explanation bottleneck models (XBMs). XBMs generate a text explanation from the input without pre-defined concepts and then predict a final task prediction based on the generated explanation by leveraging pre-trained vision-language encoder-decoder models. To achieve both the target task performance and the explanation quality, we train XBMs through the target task loss with the regularization penalizing the explanation decoder via the distillation from the frozen pre-trained decoder. Our experiments, including a comparison to state-of-the-art concept bottleneck models, confirm that XBMs provide accurate and fluent natural language explanations without pre-defined concept sets.
Scientific discovery serves as the cornerstone for advances in various fields, from the fundamental laws of physics to the intricate mechanisms of biology. However, two existing mainstream methods---symbolic regression and dimensional analysis, are significantly limited in this task: the former suffers from low computational efficiency due to the vast search space and often results in formulas without physical meaning; the latter provides a useful theoretical framework but also struggles in searching in a huge space because of lacking effective analysis for the latent variables. To address this issue, here we propose a framework for efficiently discovering underlying formulas in data, named FIND. We draw inspiration from Buckingham’s Pi theorem, imposing dimensional constraints on the input and output, thereby ensuring discovered expressions possess physical meaning. Additionally, we propose a theoretical scheme for identifying the latent structure as well as a coarse-to-fine framework, significantly reducing the search space of latent variables. This framework not only improves computational efficiency but also enhances model interpretability. From comprehensive experimental validation, FIND showcases its potential to uncover meaningful scientific insights across various domains, providing a robust tool for advancing our understanding of unknown systems.
Graph Neural Networks are powerful tools for modeling graph-structured data but their interpretability remains a significant challenge. Existing model-agnostic GNN explainers aim to identify critical subgraphs or node features relevant to task predictions but often rely on GNN predictions for supervision, lacking ground-truth explanations. This limitation can introduce biases, causing explanations to fail in accurately reflecting the GNN's decision-making processes. To address this, we propose a novel explainer for GNNs with graph segmentation and contrastive learning. Our model introduces a graph segmentation learning module to divide the input graph into explanatory and redundant subgraphs. Next, we implement edge perturbation to augment these subgraphs, generating multiple positive and negative pairs for contrastive learning between explanatory and redundant subgraphs. Finally, we develop a contrastive learning module to guide the learning of explanatory and redundant subgraphs by pulling positive pairs with the same explanatory subgraphs closer while pushing negative pairs with different explanatory subgraphs far away. This approach allows for a clearer distinction of critical subgraphs, enhancing the fidelity of the explanations. We conducted extensive experiments on graph classification and node classification tasks, demonstrating the effectiveness of the proposed method.
An event sequence generated by a temporal point process is often associated with a hidden and structured event branching process that captures the triggering relations between its historical and current events. In this study, we design a new plug-and-play module based on the Bregman ADMM (BADMM) algorithm, which infers event branches associated with event sequences in the maximum likelihood estimation framework of temporal point processes (TPPs). Specifically, we formulate the inference of event branches as an optimization problem of event transition matrix under sparse and low-rank constraints, which is embedded in existing TPP models or their learning paradigms. We can implement this optimization problem based on subspace clustering and sparse group-lasso, respectively, and solve it using the Bregman ADMM algorithm, whose unrolling leads to the proposed BADMM module. When learning a classic TPP (e.g., Hawkes process) by the expectation-maximization algorithm, the BADMM module helps derive structured responsibility matrices in the E-step. Similarly, the BADMM module helps derive low-rank and sparse attention maps for the neural TPPs with self-attention layers. The structured responsibility matrices and attention maps, which work as learned event transition matrices, indicate event branches, e.g., inferring isolated events and those key events triggering many subsequent events. Experiments on both synthetic and real-world data show that plugging our BADMM module into existing TPP models and learning paradigms can improve model performance and provide us with interpretable structured event branches.
Graph Neural Networks (GNNs) have shown efficacy in graph node classification, but face computational challenges on large-scale graphs. Although existing graph reduction methods address these issues, they still require high computational resources and fail to prioritize robust performance on out-of-distribution data. To tackle these challenges, we introduce the subgraph invariant learning paradigm, inspired by the small-world phenomenon. This approach enables models trained on specific subgraphs to generalize across diverse subgraphs, reducing computational demands, and enhancing scalability. To promote generalization, we maximize the invariance log-likelihood by deriving a theoretical lower bound of it and formulating the InVar loss. This loss minimizes the discrepancy between node representations and their corresponding invariance representations while maximizing the entropy of the node representation. In response to InVar loss, we propose the Invariance Facilitation Model (IFM), comprising the Invariance Representation Encoder (IRE) and Node Representation Encoder (NRE). IRE, capturing the invariance representations, utilizes Invariance ATTention (InvarATT) to compress long-range dependencies, while NRE learns the node representation, by integrating invariance representations via Telematic ATTention (TeleATT) and exchanging local information within each subgraph through GNNs. Evaluations on four large-scale graph datasets demonstrate the effectiveness, computational efficiency, and interpretability of IFM for large-scale graph node classification.