Prompt engineering is pivotal for harnessing the capabilities of large language models (LLMs) across diverse applications. While existing prompt optimization methods improve prompt effectiveness, they often lead to prompt drifting, where newly generated prompts can adversely impact previously successful cases while addressing failures. Furthermore, these methods tend to rely heavily on LLMs' intrinsic capabilities for prompt optimization tasks. In this paper, we introduce StraGo (Strategic-Guided Optimization), a novel approach designed to mitigate prompt drifting by leveraging insights from both successful and failed cases to identify critical factors for achieving optimization objectives. StraGo employs a how-to-do methodology, integrating in-context learning to formulate specific, actionable strategies that provide detailed, step-by-step guidance for prompt optimization. Extensive experiments conducted across a range of tasks, including reasoning, natural language understanding, domain-specific knowledge, and industrial applications, demonstrate StraGo's superior performance. It establishes a new state-of-the-art in prompt optimization, showcasing its ability to deliver stable and effective prompt improvements.
This paper introduces Knowledge Representation Augmented Generation (KRAG), a novel framework designed to enhance the capabilities of Large Language Models (LLMs) within domain-specific applications. KRAG points to the strategic inclusion of critical knowledge entities and relationships that are typically absent in standard data sets and which LLMs do not inherently learn. In the context of legal applications, we present Soft PROLEG, an implementation model under KRAG, which uses inference graphs to aid LLMs in delivering structured legal reasoning, argumentation, and explanations tailored to user inquiries. The integration of KRAG, either as a standalone framework or in tandem with retrieval augmented generation (RAG), markedly improves the ability of language models to navigate and solve the intricate challenges posed by legal texts and terminologies. This paper details KRAG's methodology, its implementation through Soft PROLEG, and potential broader applications, underscoring its significant role in advancing natural language understanding and processing in specialized knowledge domains.
Compressive adaptation approaches, such as QLoRA, are widely popular alternatives for reducing memory requirements during fine-tuning of large language models (LLMs) while producing models capable of handling various downstream tasks. The key idea is to employ a "two-tower" architecture: compressing pre-trained LLM parameters into compact representations and fine-tuning the additive full-precision adapter, which typically has few tunable parameters in low-rank format. However, the strict algebraic assumptions, such as low-rank assumption, and the complexity of composing two-tower architectures are some of the known shortcomings, resulting in a poor accuracy-efficiency trade-off. In response to these known limitations, we propose SpaLLM (Sketched Parameter Adaptation of LLMs), a novel compressive adaptation approach for LLMs. This method is also the first to illustrate parameter-sharing compression methods for LLM fine-tuning, which, unlike QLoRA, are free from strict low-rank algebraic assumptions on adapters. Furthermore, our proposal unifies model compression and adaptation into a single, streamlined process, eliminating the need for two-tower architectures. SpaLLM sketches pre-trained LLM weights into lookup tables and directly fine-tunes the values in these tables. This approach simplifies LLMs' compressive adaptation workflow, potentially improves multi-user serving efficiency, and delivers significantly better accuracy for both natural language understanding and generation tasks. Moreover, by avoiding the "two-tower" architecture, our framework only requires one compressed matrix multiplication per layer during inference, demonstrating superior inference efficiency compared to previous methods.
Software engineering (SE) chatbots are increasingly gaining attention for their role in enhancing development processes. At the core of chatbots are the Natural Language Understanding platforms (NLUs), which enable them to comprehend and respond to user queries. Before deploying NLUs, there is a need to train them with labeled data. However, acquiring such labeled data for SE chatbots is challenging due to the scarcity of high-quality datasets. This challenge arises because training SE chatbots requires specialized vocabulary and phrases not found in typical language datasets. Consequently, chatbot developers often resort to manually annotating user queries to gather the data necessary for training effective chatbots, a process that is both time-consuming and resource-intensive. Previous studies propose approaches to support chatbot practitioners in annotating users' posed queries. However, these approaches require human intervention to generate rules, called labeling functions (LFs), that identify and categorize user queries based on specific patterns in the data. To address this issue, we propose an approach to automatically generate LFs by extracting patterns from labeled user queries. We evaluate the effectiveness of our approach by applying it to the queries of four diverse SE datasets (namely AskGit, MSA, Ask Ubuntu, and Stack Overflow) and measure the performance improvement gained from training the NLU on the queries labeled by the generated LFs. We find that the generated LFs effectively label data with AUC scores of up to 85.3%, and NLU's performance improvement of up to 27.2% across the studied datasets. Furthermore, our results show that the number of LFs used to generate LFs affects the labeling performance. We believe that our approach can save time and resources in labeling users' queries, allowing practitioners to focus on core chatbot functionalities.
The development of Long-Context Large Language Models (LLMs) has markedly advanced natural language processing by facilitating the process of textual data across long documents and multiple corpora. However, Long-Context LLMs still face two critical challenges: The lost in the middle phenomenon, where crucial middle-context information is likely to be missed, and the distraction issue that the models lose focus due to overly extended contexts. To address these challenges, we propose the Context Filtering Language Model (FltLM), a novel integrated Long-Context LLM which enhances the ability of the model on multi-document question-answering (QA) tasks. Specifically, FltLM innovatively incorporates a context filter with a soft mask mechanism, identifying and dynamically excluding irrelevant content to concentrate on pertinent information for better comprehension and reasoning. Our approach not only mitigates these two challenges, but also enables the model to operate conveniently in a single forward pass. Experimental results demonstrate that FltLM significantly outperforms supervised fine-tuning and retrieval-based methods in complex QA scenarios, suggesting a promising solution for more accurate and reliable long-context natural language understanding applications.
To mitigate societal biases implicitly encoded in recent successful pretrained language models, a diverse array of approaches have been proposed to encourage model fairness, focusing on prompting, data augmentation, regularized fine-tuning, and more. Despite the development, it is nontrivial to reach a principled understanding of fairness and an effective algorithm that can consistently debias language models. In this work, by rigorous evaluations of Neural Collapse -- a learning phenomenon happen in last-layer representations and classifiers in deep networks -- on fairness-related words, we find that debiased language models exhibit collapsed alignment between token representations and word embeddings. More importantly, this observation inspires us to design a principled fine-tuning method that can effectively improve fairness in a wide range of debiasing methods, while still preserving the performance of language models on standard natural language understanding tasks. We attach our code at https://github.com/Xujxyang/Fairness-NC-main.
Large Language Models (LLMs) have demonstrated remarkable capabilities across various fields, from natural language understanding to text generation. Compared to non-generative LLMs like BERT and DeBERTa, generative LLMs like GPT series and Llama series are currently the main focus due to their superior algorithmic performance. The advancements in generative LLMs are closely intertwined with the development of hardware capabilities. Various hardware platforms exhibit distinct hardware characteristics, which can help improve LLM inference performance. Therefore, this paper comprehensively surveys efficient generative LLM inference on different hardware platforms. First, we provide an overview of the algorithm architecture of mainstream generative LLMs and delve into the inference process. Then, we summarize different optimization methods for different platforms such as CPU, GPU, FPGA, ASIC, and PIM/NDP, and provide inference results for generative LLMs. Furthermore, we perform a qualitative and quantitative comparison of inference performance with batch sizes 1 and 8 on different hardware platforms by considering hardware power consumption, absolute inference speed (tokens/s), and energy efficiency (tokens/J). We compare the performance of the same optimization methods across different hardware platforms, the performance across different hardware platforms, and the performance of different methods on the same hardware platform. This provides a systematic and comprehensive summary of existing inference acceleration work by integrating software optimization methods and hardware platforms. We point out that three trends (multimodality, inference-time compute, and higher inference energy efficiency) are promising to redefine the capabilities of edge artificial intelligence systems. Our project is available at https://dai.sjtu.edu.cn/project.html.
Low Rank Adaptation (LoRA) is a popular Parameter Efficient Fine Tuning (PEFT) method that effectively adapts large pre-trained models for downstream tasks. LoRA parameterizes model updates using low-rank matrices at each layer, significantly reducing the number of trainable parameters and, consequently, resource requirements during fine-tuning. However, the lower bound on the number of trainable parameters remains high due to the use of the low-rank matrix model. Recent works have addressed this limitation by proposing low rank tensor parameterizations for model updates. However, they only exploit redundancy across layers, or tensorize individual matrices using ad-hoc schemes that introduce additional hyperparameters. In this work, we propose a higher-order Candecomp/Parafac (CP) decomposition, enabling a more compact and flexible representation compared to existing matrix and tensor based PEFT methods. Our experiments on Natural Language Understanding, Instruction Tuning, Preference Optimization and Protein Folding benchmarks demonstrate that our method can achieve a reduction in the number of parameters while maintaining comparable performance.
Grammar induction has made significant progress in recent years. However, it is not clear how the application of induced grammar could enhance practical performance in downstream tasks. In this work, we introduce an unsupervised grammar induction method for language understanding and generation. We construct a grammar parser to induce constituency structures and dependency relations, which is simultaneously trained on downstream tasks without additional syntax annotations. The induced grammar features are subsequently incorporated into Transformer as a syntactic mask to guide self-attention. We evaluate and apply our method to multiple machine translation tasks and natural language understanding tasks. Our method demonstrates superior performance compared to the original Transformer and other models enhanced with external parsers. Experimental results indicate that our method is effective in both from-scratch and pre-trained scenarios. Additionally, our research highlights the contribution of explicitly modeling the grammatical structure of texts to neural network models.
Large language models (LLMs) have made significant progress in natural language understanding and generation, driven by scalable pretraining and advanced finetuning. However, enhancing reasoning abilities in LLMs, particularly via reinforcement learning from human feedback (RLHF), remains challenging due to the scarcity of high-quality preference data, which is labor-intensive to annotate and crucial for reward model (RM) finetuning. To alleviate this issue, we introduce CodePMP, a scalable preference model pretraining (PMP) pipeline that utilizes a large corpus of synthesized code-preference pairs from publicly available high-quality source code. CodePMP improves RM finetuning efficiency by pretraining preference models on large-scale synthesized code-preference pairs. We evaluate CodePMP on mathematical reasoning tasks (GSM8K, MATH) and logical reasoning tasks (ReClor, LogiQA2.0), consistently showing significant improvements in reasoning performance of LLMs and highlighting the importance of scalable preference model pretraining for efficient reward modeling.
Large neural networks spend most computation on floating point tensor multiplications. In this work, we find that a floating point multiplier can be approximated by one integer adder with high precision. We propose the linear-complexity multiplication L-Mul algorithm that approximates floating point number multiplication with integer addition operations. The new algorithm costs significantly less computation resource than 8-bit floating point multiplication but achieves higher precision. Compared to 8-bit floating point multiplications, the proposed method achieves higher precision but consumes significantly less bit-level computation. Since multiplying floating point numbers requires substantially higher energy compared to integer addition operations, applying the L-Mul operation in tensor processing hardware can potentially reduce 95% energy cost by element-wise floating point tensor multiplications and 80% energy cost of dot products. We calculated the theoretical error expectation of L-Mul, and evaluated the algorithm on a wide range of textual, visual, and symbolic tasks, including natural language understanding, structural reasoning, mathematics, and commonsense question answering. Our numerical analysis experiments agree with the theoretical error estimation, which indicates that L-Mul with 4-bit mantissa achieves comparable precision as float8_e4m3 multiplications, and L-Mul with 3-bit mantissa outperforms float8_e5m2. Evaluation results on popular benchmarks show that directly applying L-Mul to the attention mechanism is almost lossless. We further show that replacing all floating point multiplications with 3-bit mantissa L-Mul in a transformer model achieves equivalent precision as using float8_e4m3 as accumulation precision in both fine-tuning and inference.
We investigate LoRA in federated learning through the lens of the asymmetry analysis of the learned $A$ and $B$ matrices. In doing so, we uncover that $A$ matrices are responsible for learning general knowledge, while $B$ matrices focus on capturing client-specific knowledge. Based on this finding, we introduce Federated Share-A Low-Rank Adaptation (FedSA-LoRA), which employs two low-rank trainable matrices $A$ and $B$ to model the weight update, but only $A$ matrices are shared with the server for aggregation. Moreover, we delve into the relationship between the learned $A$ and $B$ matrices in other LoRA variants, such as rsLoRA and VeRA, revealing a consistent pattern. Consequently, we extend our FedSA-LoRA method to these LoRA variants, resulting in FedSA-rsLoRA and FedSA-VeRA. In this way, we establish a general paradigm for integrating LoRA with FL, offering guidance for future work on subsequent LoRA variants combined with FL. Extensive experimental results on natural language understanding and generation tasks demonstrate the effectiveness of the proposed method. Our code is available at https://github.com/Pengxin-Guo/FedSA-LoRA.
Integrating machine learning (ML) into customer service chatbots enhances their ability to understand and respond to user queries, ultimately improving service performance. However, they may appear artificial to some users and affecting customer experience. Hence, meticulous evaluation of ML models for each pipeline component is crucial for optimizing performance, though differences in functionalities can lead to unfair comparisons. In this paper, we present a tailored experimental evaluation approach for goal-oriented customer service chatbots with pipeline architecture, focusing on three key components: Natural Language Understanding (NLU), dialogue management (DM), and Natural Language Generation (NLG). Our methodology emphasizes individual assessment to determine optimal ML models. Specifically, we focus on optimizing hyperparameters and evaluating candidate models for NLU (utilizing BERT and LSTM), DM (employing DQN and DDQN), and NLG (leveraging GPT-2 and DialoGPT). The results show that for the NLU component, BERT excelled in intent detection whereas LSTM was superior for slot filling. For the DM component, the DDQN model outperformed DQN by achieving fewer turns, higher rewards, as well as greater success rates. For NLG, the large language model GPT-2 surpassed DialoGPT in BLEU, METEOR, and ROUGE metrics. These findings aim to provide a benchmark for future research in developing and optimizing customer service chatbots, offering valuable insights into model performance and optimal hyperparameters.
Large Language Models (LLMs) have been profusely evaluated on their ability to answer questions on many topics and their performance on different natural language understanding tasks. Those tests are usually conducted in English, but most LLM users are not native English speakers. Therefore, it is of interest to analyze how LLMs understand other languages at different levels: from paragraphs to morphems. In this paper, we evaluate the performance of state-of-the-art LLMs in TELEIA, a recently released benchmark with similar questions to those of Spanish exams for foreign students, covering topics such as reading comprehension, word formation, meaning and compositional semantics, and grammar. The results show that LLMs perform well at understanding Spanish but are still far from achieving the level of a native speaker in terms of grammatical competence.
Vision language models have played a key role in extracting meaningful features for various robotic applications. Among these, Contrastive Language-Image Pretraining (CLIP) is widely used in robotic tasks that require both vision and natural language understanding. However, CLIP was trained solely on static images paired with text prompts and has not yet been fully adapted for robotic tasks involving dynamic actions. In this paper, we introduce Robotic-CLIP to enhance robotic perception capabilities. We first gather and label large-scale action data, and then build our Robotic-CLIP by fine-tuning CLIP on 309,433 videos (~7.4 million frames) of action data using contrastive learning. By leveraging action data, Robotic-CLIP inherits CLIP's strong image performance while gaining the ability to understand actions in robotic contexts. Intensive experiments show that our Robotic-CLIP outperforms other CLIP-based models across various language-driven robotic tasks. Additionally, we demonstrate the practical effectiveness of Robotic-CLIP in real-world grasping applications.
Parameter-Efficient Fine-Tuning (PEFT) methods, particularly Low-Rank Adaptation (LoRA), effectively reduce the number of trainable parameters in Large Language Models (LLMs). However, as model scales continue to grow, the demand for computational resources remains a significant challenge. Existing LoRA variants often struggle to strike an optimal balance between adaptability (model performance and convergence speed) and efficiency (computational overhead, memory usage, and initialization time). This paper introduces FOSSIL(\textbf{F}ramework for \textbf{O}ptimal \textbf{S}hard \textbf{S}haring \textbf{I}ntegration in \textbf{L}oRA), a novel PEFT approach that addresses this trade-off through a simple shard-sharing mechanism. FOSSIL leverages the insight that a low-rank adaptation can be achieved by decomposing the weight matrix into multiple fragment matrices and utilizing a shared, trainable common fragment. This method constructs the low-rank update matrix through the replication of these shared, partitioned shards. We also propose a hardware-efficient and broadly applicable implementation for FOSSIL. Extensive experiments conducted on a range of tasks, alongside a systematic analysis of computational performance, demonstrate FOSSIL's superiority. The results show that FOSSIL significantly outperforms standard LoRA and its prominent variants in both model performance metrics and computational efficiency, including initialization speed and training throughput. By effectively balancing expressive power and resource utilization, FOSSIL offers a compelling solution for efficiently adapting large-scale models.
Despite the recent broad adoption of Large Language Models (LLMs) across various domains, their potential for enriching information systems in extracting and exploring Linked Data (LD) and Resource Description Framework (RDF) triplestores has not been extensively explored. This paper examines the integration of LLMs within existing systems, emphasising the enhancement of conversational user interfaces (UIs) and their capabilities for data extraction by producing more accurate SPARQL queries without the requirement for model retraining. Typically, conversational UI models necessitate retraining with the introduction of new datasets or updates, limiting their functionality as general-purpose extraction tools. Our approach addresses this limitation by incorporating LLMs into the conversational UI workflow, significantly enhancing their ability to comprehend and process user queries effectively. By leveraging the advanced natural language understanding capabilities of LLMs, our method improves RDF entity extraction within web systems employing conventional chatbots. This integration facilitates a more nuanced and context-aware interaction model, critical for handling the complex query patterns often encountered in RDF datasets and Linked Open Data (LOD) endpoints. The evaluation of this methodology shows a marked enhancement in system expressivity and the accuracy of responses to user queries, indicating a promising direction for future research in this area. This investigation not only underscores the versatility of LLMs in enhancing existing information systems but also sets the stage for further explorations into their potential applications within more specialised domains of web information systems.
Large language models (LLMs) are a promising venue for natural language understanding and generation. However, current LLMs are far from reliable: they are prone to generating non-factual information and, more crucially, to contradicting themselves when prompted to reason about relations between entities of the world. These problems are currently addressed with large scale fine-tuning or by delegating reasoning to external tools. In this work, we strive for a middle ground and introduce a loss based on neuro-symbolic reasoning that teaches an LLM to be logically consistent with an external set of facts and rules and improves self-consistency even when the LLM is fine-tuned on a limited set of facts. Our approach also allows to easily combine multiple logical constraints at once in a principled way, delivering LLMs that are more consistent w.r.t. all constraints and improve over several baselines w.r.t. a given constraint. Moreover, our method allows LLMs to extrapolate to unseen but semantically similar factual knowledge, represented in unseen datasets, more systematically.
Despite the impressive performance of autoregressive Language Models (LM) it has been shown that due to reporting bias, LMs lack visual knowledge, i.e. they do not know much about the visual world and its properties. To augment LMs with visual knowledge, existing solutions often rely on explicit images, requiring time-consuming retrieval or image generation systems. This paper shows that explicit images are not necessary to visually augment an LM. Instead, we use visually-grounded text representations obtained from the well-known CLIP multimodal system. For a fair comparison, we modify VALM, a visually-augmented LM which uses image retrieval and representation, to work directly with visually-grounded text representations. We name this new model BLIND-VALM. We show that BLIND-VALM performs on par with VALM for Visual Language Understanding (VLU), Natural Language Understanding (NLU) and Language Modeling tasks, despite being significantly more efficient and simpler. We also show that scaling up our model within the compute budget of VALM, either increasing the model or pre-training corpus size, we outperform VALM for all the evaluation tasks.
Ensuring content compliance with community guidelines is crucial for maintaining healthy online social environments. However, traditional human-based compliance checking struggles with scaling due to the increasing volume of user-generated content and a limited number of moderators. Recent advancements in Natural Language Understanding demonstrated by Large Language Models unlock new opportunities for automated content compliance verification. This work evaluates six AI-agents built on Open-LLMs for automated rule compliance checking in Decentralized Social Networks, a challenging environment due to heterogeneous community scopes and rules. Analyzing over 50,000 posts from hundreds of Mastodon servers, we find that AI-agents effectively detect non-compliant content, grasp linguistic subtleties, and adapt to diverse community contexts. Most agents also show high inter-rater reliability and consistency in score justification and suggestions for compliance. Human-based evaluation with domain experts confirmed the agents' reliability and usefulness, rendering them promising tools for semi-automated or human-in-the-loop content moderation systems.