Large language models (LLMs) have exhibited an array of reasoning capabilities but face challenges like error propagation and hallucination, particularly in specialised areas like finance, where data is heterogeneous, and precision is paramount. We explore the potential of language model augmentation with external tools to mitigate these limitations and offload certain reasoning steps to external tools that are more suited for the task, instead of solely depending on the LLM's inherent abilities. More concretely, using financial domain question-answering datasets, we apply supervised fine-tuning on a LLaMA-2 13B Chat model to act both as a 'task router' and 'task solver'. The 'task router' dynamically directs a question to either be answered internally by the LLM or externally via the right tool from the tool set. Our tool-equipped SFT model, Raven, demonstrates an improvement of 35.2% and 5.06% over the base model and SFT-only baselines, respectively, and is highly competitive with strong GPT-3.5 results. To the best of our knowledge, our work is the first that investigates tool augmentation of language models for the finance domain.
相關內容
Gaussian processes (GPs) are commonly used for geospatial analysis, but they suffer from high computational complexity when dealing with massive data. For instance, the log-likelihood function required in estimating the statistical model parameters for geospatial data is a computationally intensive procedure that involves computing the inverse of a covariance matrix with size n X n, where n represents the number of geographical locations. As a result, in the literature, studies have shifted towards approximation methods to handle larger values of n effectively while maintaining high accuracy. These methods encompass a range of techniques, including low-rank and sparse approximations. Vecchia approximation is one of the most promising methods to speed up evaluating the log-likelihood function. This study presents a parallel implementation of the Vecchia approximation, utilizing batched matrix computations on contemporary GPUs. The proposed implementation relies on batched linear algebra routines to efficiently execute individual conditional distributions in the Vecchia algorithm. We rely on the KBLAS linear algebra library to perform batched linear algebra operations, reducing the time to solution compared to the state-of-the-art parallel implementation of the likelihood estimation operation in the ExaGeoStat software by up to 700X, 833X, 1380X on 32GB GV100, 80GB A100, and 80GB H100 GPUs, respectively. We also successfully manage larger problem sizes on a single NVIDIA GPU, accommodating up to 1M locations with 80GB A100 and H100 GPUs while maintaining the necessary application accuracy. We further assess the accuracy performance of the implemented algorithm, identifying the optimal settings for the Vecchia approximation algorithm to preserve accuracy on two real geospatial datasets: soil moisture data in the Mississippi Basin area and wind speed data in the Middle East.
Foundation models, such as Large language Models (LLMs), have attracted significant amount of interest due to their large number of applications. Existing works show that appropriate prompt design, such as Chain-of-Thoughts, can unlock LLM's powerful capacity in diverse areas. However, when handling tasks involving repetitive sub-tasks and/or deceptive contents, such as arithmetic calculation and article-level fake news detection, existing prompting strategies either suffers from insufficient expressive power or intermediate errors triggered by hallucination. To make LLM more discerning to such intermediate errors, we propose to guide LLM with a Divide-and-Conquer program that simultaneously ensures superior expressive power and disentangles task decomposition, sub-task resolution, and resolution assembly process. Theoretic analysis reveals that our strategy can guide LLM to extend the expressive power of fixed-depth Transformer. Experiments indicate that our proposed method can achieve better performance than typical prompting strategies in tasks bothered by intermediate errors and deceptive contents, such as large integer multiplication, hallucination detection and misinformation detection.
The recent advances in language-based generative models have paved the way for the orchestration of multiple generators of different artefact types (text, image, audio, etc.) into one system. Presently, many open-source pre-trained models combine text with other modalities, thus enabling shared vector embeddings to be compared across different generators. Within this context we propose a novel approach to handle multimodal creative tasks using Quality Diversity evolution. Our contribution is a variation of the MAP-Elites algorithm, MAP-Elites with Transverse Assessment (MEliTA), which is tailored for multimodal creative tasks and leverages deep learned models that assess coherence across modalities. MEliTA decouples the artefacts' modalities and promotes cross-pollination between elites. As a test bed for this algorithm, we generate text descriptions and cover images for a hypothetical video game and assign each artefact a unique modality-specific behavioural characteristic. Results indicate that MEliTA can improve text-to-image mappings within the solution space, compared to a baseline MAP-Elites algorithm that strictly treats each image-text pair as one solution. Our approach represents a significant step forward in multimodal bottom-up orchestration and lays the groundwork for more complex systems coordinating multimodal creative agents in the future.
Despite their impressive capabilities, large language models (LLMs) have been observed to generate responses that include inaccurate or fabricated information, a phenomenon commonly known as ``hallucination''. In this work, we propose a simple \textit{Induce-then-Contrast} Decoding (ICD) strategy to alleviate hallucinations. We first construct a factually weak LLM by inducing hallucinations from the original LLMs. Then, we penalize these induced hallucinations during decoding to enhance the factuality of the generated content. Concretely, we determine the final next-token predictions by amplifying the predictions from the original model and downplaying the induced untruthful predictions via contrastive decoding. Experimental results on both discrimination-based and generation-based hallucination evaluation benchmarks, such as TruthfulQA and \textsc{FActScore}, demonstrate that our proposed ICD methods can effectively enhance the factuality of LLMs across various model sizes and families. For example, when equipped with ICD, Llama2-7B-Chat and Mistral-7B-Instruct achieve performance comparable to ChatGPT and GPT4 on TruthfulQA, respectively.
Large language models (LLMs) have demonstrated remarkable capabilities across various NLP tasks. However, their computational costs are prohibitively high. To address this issue, previous research has attempted to distill the knowledge of LLMs into smaller models by generating annotated data. Nonetheless, these works have mainly focused on the direct use of LLMs for text generation and labeling, without fully exploring their potential to comprehend the target task and acquire valuable knowledge. In this paper, we propose EvoKD: Evolving Knowledge Distillation, which leverages the concept of active learning to interactively enhance the process of data generation using large language models, simultaneously improving the task capabilities of small domain model (student model). Different from previous work, we actively analyze the student model's weaknesses, and then synthesize labeled samples based on the analysis. In addition, we provide iterative feedback to the LLMs regarding the student model's performance to continuously construct diversified and challenging samples. Experiments and analysis on different NLP tasks, namely, text classification and named entity recognition show the effectiveness of EvoKD.
User alignment is crucial for adapting general-purpose language models (LMs) to downstream tasks, but human annotations are often not available for all types of instructions, especially those with customized constraints. We observe that user instructions typically contain constraints. While assessing response quality in terms of the whole instruction is often costly, efficiently evaluating the satisfaction rate of constraints is feasible. We investigate common constraints in NLP tasks, categorize them into three classes based on the types of their arguments, and propose a unified framework, ACT (Aligning to ConsTraints), to automatically produce supervision signals for user alignment with constraints. Specifically, ACT uses constraint verifiers, which are typically easy to implement in practice, to compute constraint satisfaction rate (CSR) of each response. It samples multiple responses for each prompt and collect preference labels based on their CSR automatically. Subsequently, ACT adapts the LM to the target task through a ranking-based learning process. Experiments on fine-grained entity typing, abstractive summarization, and temporal question answering show that ACT is able to enhance LMs' capability to adhere to different classes of constraints, thereby improving task performance. Further experiments show that the constraint-following capabilities are transferable.
Machine learning models are commonly tested in-distribution (same dataset); performance almost always drops in out-of-distribution settings. For HRI research, the goal is often to develop generalized models. This makes domain generalization - retaining performance in different settings - a critical issue. In this study, we present a concise analysis of domain generalization in failure detection models trained on human facial expressions. Using two distinct datasets of humans reacting to videos where error occurs, one from a controlled lab setting and another collected online, we trained deep learning models on each dataset. When testing these models on the alternate dataset, we observed a significant performance drop. We reflect on the causes for the observed model behavior and leave recommendations. This work emphasizes the need for HRI research focusing on improving model robustness and real-life applicability.
Although large language models (LLMs) are widely deployed, the data used to train them is rarely disclosed. Given the incredible scale of this data, up to trillions of tokens, it is all but certain that it includes potentially problematic text such as copyrighted materials, personally identifiable information, and test data for widely reported reference benchmarks. However, we currently have no way to know which data of these types is included or in what proportions. In this paper, we study the pretraining data detection problem: given a piece of text and black-box access to an LLM without knowing the pretraining data, can we determine if the model was trained on the provided text? To facilitate this study, we introduce a dynamic benchmark WIKIMIA that uses data created before and after model training to support gold truth detection. We also introduce a new detection method Min-K% Prob based on a simple hypothesis: an unseen example is likely to contain a few outlier words with low probabilities under the LLM, while a seen example is less likely to have words with such low probabilities. Min-K% Prob can be applied without any knowledge about the pretraining corpus or any additional training, departing from previous detection methods that require training a reference model on data that is similar to the pretraining data. Moreover, our experiments demonstrate that Min-K% Prob achieves a 7.4% improvement on WIKIMIA over these previous methods. We apply Min-K% Prob to three real-world scenarios, copyrighted book detection, contaminated downstream example detection and privacy auditing of machine unlearning, and find it a consistently effective solution.
Current natural language understanding (NLU) models have been continuously scaling up, both in terms of model size and input context, introducing more hidden and input neurons. While this generally improves performance on average, the extra neurons do not yield a consistent improvement for all instances. This is because some hidden neurons are redundant, and the noise mixed in input neurons tends to distract the model. Previous work mainly focuses on extrinsically reducing low-utility neurons by additional post- or pre-processing, such as network pruning and context selection, to avoid this problem. Beyond that, can we make the model reduce redundant parameters and suppress input noise by intrinsically enhancing the utility of each neuron? If a model can efficiently utilize neurons, no matter which neurons are ablated (disabled), the ablated submodel should perform no better than the original full model. Based on such a comparison principle between models, we propose a cross-model comparative loss for a broad range of tasks. Comparative loss is essentially a ranking loss on top of the task-specific losses of the full and ablated models, with the expectation that the task-specific loss of the full model is minimal. We demonstrate the universal effectiveness of comparative loss through extensive experiments on 14 datasets from 3 distinct NLU tasks based on 5 widely used pretrained language models and find it particularly superior for models with few parameters or long input.
Visual Question Answering (VQA) models have struggled with counting objects in natural images so far. We identify a fundamental problem due to soft attention in these models as a cause. To circumvent this problem, we propose a neural network component that allows robust counting from object proposals. Experiments on a toy task show the effectiveness of this component and we obtain state-of-the-art accuracy on the number category of the VQA v2 dataset without negatively affecting other categories, even outperforming ensemble models with our single model. On a difficult balanced pair metric, the component gives a substantial improvement in counting over a strong baseline by 6.6%.
小貼士
登錄享
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191