Artificial intelligence is making significant strides in the finance industry, revolutionizing how data is processed and interpreted. Among these technologies, large language models (LLMs) have demonstrated substantial potential to transform financial services by automating complex tasks, enhancing customer service, and providing detailed financial analysis. Firstly, we introduce IDEA-FinBench, an evaluation benchmark specifically tailored for assessing financial knowledge in large language models (LLMs). This benchmark utilizes questions from two globally respected and authoritative financial professional exams, aimimg to comprehensively evaluate the capability of LLMs to directly address exam questions pertinent to the finance sector. Secondly, we propose IDEA-FinKER, a Financial Knowledge Enhancement framework designed to facilitate the rapid adaptation of general LLMs to the financial domain, introducing a retrieval-based few-shot learning method for real-time context-level knowledge injection, and a set of high-quality financial knowledge instructions for fine-tuning any general LLM. Finally, we present IDEA-FinQA, a financial question-answering system powered by LLMs. This system is structured around a scheme of real-time knowledge injection and factual enhancement using external knowledge. IDEA-FinQA is comprised of three main modules: the data collector, the data querying module, and LLM-based agents tasked with specific functions.

Large Language Models (LLMs) excel in fluency but risk producing inaccurate content, called "hallucinations." This paper outlines a standardized process for categorizing fine-grained hallucination types and proposes an innovative framework--the Progressive Fine-grained Model Editor (PFME)--specifically designed to detect and correct fine-grained hallucinations in LLMs. PFME consists of two collaborative modules: the Real-time Fact Retrieval Module and the Fine-grained Hallucination Detection and Editing Module. The former identifies key entities in the document and retrieves the latest factual evidence from credible sources. The latter further segments the document into sentence-level text and, based on relevant evidence and previously edited context, identifies, locates, and edits each sentence's hallucination type. Experimental results on FavaBench and FActScore demonstrate that PFME outperforms existing methods in fine-grained hallucination detection tasks. Particularly, when using the Llama3-8B-Instruct model, PFME's performance in fine-grained hallucination detection with external knowledge assistance improves by 8.7 percentage points (pp) compared to ChatGPT. In editing tasks, PFME further enhances the FActScore of FActScore-Alpaca13B and FActScore-ChatGPT datasets, increasing by 16.2pp and 4.6pp, respectively.

Uncertainty quantification (UQ) in natural language generation (NLG) tasks remains an open challenge, exacerbated by the intricate nature of the recent large language models (LLMs). This study investigates adapting conformal prediction (CP), which can convert any heuristic measure of uncertainty into rigorous theoretical guarantees by constructing prediction sets, for black-box LLMs in open-ended NLG tasks. We propose a sampling-based uncertainty measure leveraging self-consistency and develop a conformal uncertainty criterion by integrating the uncertainty condition aligned with correctness into the design of the CP algorithm. Experimental results indicate that our uncertainty measure generally surpasses prior state-of-the-art methods. Furthermore, we calibrate the prediction sets within the model's unfixed answer distribution and achieve strict control over the correctness coverage rate across 6 LLMs on 4 free-form NLG datasets, spanning general-purpose and medical domains, while the small average set size further highlights the efficiency of our method in providing trustworthy guarantees for practical open-ended NLG applications.

Large language model (LLM)-based applications consist of both LLM and non-LLM components, each contributing to the end-to-end latency. Despite great efforts to optimize LLM inference, end-to-end workflow optimization has been overlooked. Existing frameworks employ coarse-grained orchestration with task modules, which confines optimizations to within each module and yields suboptimal scheduling decisions. We propose fine-grained end-to-end orchestration, which utilizes task primitives as the basic units and represents each query's workflow as a primitive-level dataflow graph. This explicitly exposes a much larger design space, enables optimizations in parallelization and pipelining across primitives of different modules, and enhances scheduling to improve application-level performance. We build Teola, a novel orchestration framework for LLM-based applications that implements this scheme. Comprehensive experiments show that Teola can achieve up to 2.09x speedup over existing systems across various popular LLM applications.

We investigate the use of Large Language Models (LLMs) to equip neural robotic agents with human-like social and cognitive competencies, for the purpose of open-ended human-robot conversation and collaboration. We introduce a modular and extensible methodology for grounding an LLM with the sensory perceptions and capabilities of a physical robot, and integrate multiple deep learning models throughout the architecture in a form of system integration. The integrated models encompass various functions such as speech recognition, speech generation, open-vocabulary object detection, human pose estimation, and gesture detection, with the LLM serving as the central text-based coordinating unit. The qualitative and quantitative results demonstrate the huge potential of LLMs in providing emergent cognition and interactive language-oriented control of robots in a natural and social manner.

Large language models (LLMs) have brought a paradigm shift to the field of code generation, offering the potential to enhance the software development process. However, previous research mainly focuses on the accuracy of code generation, while coding style differences between LLMs and human developers remain under-explored. In this paper, we empirically analyze the differences in coding style between the code generated by mainstream Code LLMs and the code written by human developers, and summarize coding style inconsistency taxonomy. Specifically, we first summarize the types of coding style inconsistencies by manually analyzing a large number of generation results. We then compare the code generated by Code LLMs with the code written by human programmers in terms of readability, conciseness, and robustness. The results reveal that LLMs and developers have different coding styles. Additionally, we study the possible causes of these inconsistencies and provide some solutions to alleviate the problem.

Data curation is commonly considered a "secret-sauce" for LLM training, with higher quality data usually leading to better LLM performance. Given the scale of internet-scraped corpora, data pruning has become a larger and larger focus. Specifically, many have shown that de-duplicating data, or sub-selecting higher quality data, can lead to efficiency or performance improvements. Generally, three types of methods are used to filter internetscale corpora: embedding-based, heuristic-based, and classifier-based. In this work, we contrast the former two in the domain of finetuning LLMs for code generation. We find that embedding-based methods are often confounded by length, and that a simple heuristic-- pruning long files--outperforms other methods in compute-limited regimes. Our method can yield up to a 2x efficiency benefit in training (while matching performance) or a 3.5% absolute performance improvement on HumanEval (while matching compute). However, we find that perplexity on held-out long files can increase, begging the question of whether optimizing data mixtures for common coding benchmarks (HumanEval, MBPP) actually best serves downstream use cases. Overall, we hope our work builds useful intuitions about code data (specifically, the low quality of extremely long code files) provides a compelling heuristic-based method for data pruning, and brings to light questions in how we evaluate code generation models.

Large Multimodal Models (LMMs) exhibit impressive cross-modal understanding and reasoning abilities, often assessed through multiple-choice questions (MCQs) that include an image, a question, and several options. However, many benchmarks used for such evaluations suffer from systematic biases. Remarkably, Large Language Models (LLMs) without any visual perception capabilities achieve non-trivial performance, undermining the credibility of these evaluations. To address this issue while maintaining the efficiency of MCQ evaluations, we propose MMEvalPro, a benchmark designed to avoid Type-I errors through a trilogy evaluation pipeline and more rigorous metrics. For each original question from existing benchmarks, human annotators augment it by creating one perception question and one knowledge anchor question through a meticulous annotation process. MMEvalPro comprises $2,138$ question triplets, totaling $6,414$ distinct questions. Two-thirds of these questions are manually labeled by human experts, while the rest are sourced from existing benchmarks (MMMU, ScienceQA, and MathVista). Compared with the existing benchmarks, our experiments with the latest LLMs and LMMs demonstrate that MMEvalPro is more challenging (the best LMM lags behind human performance by $31.73\%$, compared to an average gap of $8.03\%$ in previous benchmarks) and more trustworthy (the best LLM trails the best LMM by $23.09\%$, whereas the gap for previous benchmarks is just $14.64\%$). Our in-depth analysis explains the reason for the large performance gap and justifies the trustworthiness of evaluation, underscoring its significant potential for advancing future research.

Hierarchical text classification (HTC) is an important task with broad applications, while few-shot HTC has gained increasing interest recently. While in-context learning (ICL) with large language models (LLMs) has achieved significant success in few-shot learning, it is not as effective for HTC because of the expansive hierarchical label sets and extremely-ambiguous labels. In this work, we introduce the first ICL-based framework with LLM for few-shot HTC. We exploit a retrieval database to identify relevant demonstrations, and an iterative policy to manage multi-layer hierarchical labels. Particularly, we equip the retrieval database with HTC label-aware representations for the input texts, which is achieved by continual training on a pretrained language model with masked language modeling (MLM), layer-wise classification (CLS, specifically for HTC), and a novel divergent contrastive learning (DCL, mainly for adjacent semantically-similar labels) objective. Experimental results on three benchmark datasets demonstrate superior performance of our method, and we can achieve state-of-the-art results in few-shot HTC.

Each new generation of English-oriented Large Language Models (LLMs) exhibits enhanced cross-lingual transfer capabilities and significantly outperforms older LLMs on low-resource languages. This prompts the question: Is there a need for LLMs dedicated to a particular low-resource language? We aim to explore this question for Bengali, a low-to-moderate resource Indo-Aryan language native to the Bengal region of South Asia. We compare the performance of open-weight and closed-source LLMs such as LLaMA-3 and GPT-4 against fine-tuned encoder-decoder models across a diverse set of Bengali downstream tasks, including translation, summarization, paraphrasing, question-answering, and natural language inference. Our findings reveal that while LLMs generally excel in reasoning tasks, their performance in tasks requiring Bengali script generation is inconsistent. Key challenges include inefficient tokenization of Bengali script by existing LLMs, leading to increased computational costs and potential performance degradation. Additionally, we highlight biases in machine-translated datasets commonly used for Bengali NLP tasks. We conclude that there is a significant need for a Bengali-oriented LLM, but the field currently lacks the high-quality pretraining and instruction-tuning datasets necessary to develop a highly effective model.