Gen-AI techniques are able to improve understanding of context and nuances in language modeling, translation between languages, handle large volumes of data, provide fast, low-latency responses and can be fine-tuned for various tasks and domains. In this manuscript, we present a comprehensive overview of the applications of Gen-AI techniques in the finance domain. In particular, we present the opportunities and challenges associated with the usage of Gen-AI techniques. We also illustrate the various methodologies which can be used to train Gen-AI techniques and present the various application areas of Gen-AI technologies in the finance ecosystem. To the best of our knowledge, this work represents the most comprehensive summarization of Gen-AI techniques within the financial domain. The analysis is designed for a deep overview of areas marked for substantial advancement while simultaneously pin-point those warranting future prioritization. We also hope that this work would serve as a conduit between finance and other domains, thus fostering the cross-pollination of innovative concepts and practices.
Reward models (RMs) have driven the state-of-the-art performance of LLMs today by enabling the integration of human feedback into the language modeling process. However, RMs are primarily trained and evaluated in English, and their capabilities in multilingual settings remain largely understudied. In this work, we conduct a systematic evaluation of several reward models in multilingual settings. We first construct the first-of-its-kind multilingual RM evaluation benchmark, M-RewardBench, consisting of 2.87k preference instances for 23 typologically diverse languages, that tests the chat, safety, reasoning, and translation capabilities of RMs. We then rigorously evaluate a wide range of reward models on M-RewardBench, offering fresh insights into their performance across diverse languages. We identify a significant gap in RMs' performances between English and non-English languages and show that RM preferences can change substantially from one language to another. We also present several findings on how different multilingual aspects impact RM performance. Specifically, we show that the performance of RMs is improved with improved translation quality. Similarly, we demonstrate that the models exhibit better performance for high-resource languages. We release M-RewardBench dataset and the codebase in this study to facilitate a better understanding of RM evaluation in multilingual settings.
As large language models (LLMs) are increasingly deployed across various industries, concerns regarding their reliability, particularly due to hallucinations - outputs that are factually inaccurate or irrelevant to user input - have grown. Our research investigates the relationship between the training process and the emergence of hallucinations to address a key gap in existing research that focuses primarily on post hoc detection and mitigation strategies. Using models from the Pythia suite (70M - 12B parameters) and several hallucination detection metrics, we analyze hallucination trends throughout training and explore LLM internal dynamics. We introduce Sensitivity Dropout (SenD), a novel training protocol designed to mitigate hallucinations by reducing variance during training. SenD achieves this by deterministically dropping embedding indices with significant variability, referred to as Sensitive Embedding Indices. In addition, we develop an unsupervised hallucination detection metric, Efficient EigenScore (EES), which approximates the traditional EigenScore at 2x speed. This efficient metric is integrated into our protocol, allowing SenD to be both computationally scalable and effective at reducing hallucinations. Our empirical evaluation demonstrates that our approach improves LLM reliability at test time by up to 40% compared to normal training while also providing an efficient method to improve factual accuracy when adapting LLMs to Wikipedia, Medical, and LegalBench domains.
In the realm of Large Language Model (LLM) functionalities, providing reliable information is paramount, yet reports suggest that LLM outputs lack consistency. This inconsistency, often at-tributed to randomness in token sampling, under-mines user trust as it leads to varying responses even for identical queries. In this paper, we present a new approach for evaluating semantic consistencies of LLM including comparison of alternative tech-niques. Our approach evaluates whether LLM re-sponses are semantically congruent for a given question, recognizing that as syntactically different sentences may convey the same meaning. Here-tofore, To enhance LLM consistency, two main approaches have been explored: Leverage external knowledge as context like the RAG pattern or use Zero-shot-CoT to improve performance of LLM itself. We apply our evaluation approach to these techniques, and demonstrate to compare the im-pact of these methods on LLM response con-sistency across different domains of question an-swering tasks. Using the TruthfulQA dataset to assess LLM responses, the study induces N re-sponses per question from the LLM and clusters semantically equivalent sentences to measure semantic consistency across 37 categories. Through this, it quantitatively analyzes the effectiveness of the aforementioned methods in improving LLM performance before and after their adoption.
We present xGen-MM-Vid (BLIP-3-Video): a multimodal language model for videos, particularly designed to efficiently capture temporal information over multiple frames. BLIP-3-Video takes advantage of the 'temporal encoder' in addition to the conventional visual tokenizer, which maps a sequence of tokens over multiple frames into a compact set of visual tokens. This enables BLIP3-Video to use much fewer visual tokens than its competing models (e.g., 32 vs. 4608 tokens). We explore different types of temporal encoders, including learnable spatio-temporal pooling as well as sequential models like Token Turing Machines. We experimentally confirm that BLIP-3-Video obtains video question-answering accuracies comparable to much larger state-of-the-art models (e.g., 34B), while being much smaller (i.e., 4B) and more efficient by using fewer visual tokens. The project website is at https://www.salesforceairesearch.com/opensource/xGen-MM-Vid/index.html
Chain-of-thought (CoT) reasoning in vision language models (VLMs) is crucial for improving interpretability and trustworthiness. However, current training recipes lack robust CoT reasoning data, relying on datasets dominated by short annotations with minimal rationales. In this work, we show that training VLM on short answers does not generalize well to reasoning tasks that require more detailed responses. To address this, we propose a two-fold approach. First, we distill rationales from GPT-4o model to enrich the training data and fine-tune VLMs, boosting their CoT performance. Second, we apply reinforcement learning to further calibrate reasoning quality. Specifically, we construct positive (correct) and negative (incorrect) pairs of model-generated reasoning chains, by comparing their predictions with annotated short answers. Using this pairwise data, we apply the Direct Preference Optimization algorithm to refine the model's reasoning abilities. Our experiments demonstrate significant improvements in CoT reasoning on benchmark datasets and better generalization to direct answer prediction as well. This work emphasizes the importance of incorporating detailed rationales in training and leveraging reinforcement learning to strengthen the reasoning capabilities of VLMs.
Some information is factual (e.g., "Paris is in France"), whereas other information is probabilistic (e.g., "the coin flip will be a [Heads/Tails]."). We believe that good Language Models (LMs) should understand and reflect this nuance. Our work investigates this by testing if LMs' output probabilities are calibrated to their textual contexts. We define model "calibration" as the degree to which the output probabilities of candidate tokens are aligned with the relative likelihood that should be inferred from the given context. For example, if the context concerns two equally likely options (e.g., heads or tails for a fair coin), the output probabilities should reflect this. Likewise, context that concerns non-uniformly likely events (e.g., rolling a six with a die) should also be appropriately captured with proportionate output probabilities. We find that even in simple settings the best LMs (1) are poorly calibrated, and (2) have systematic biases (e.g., preferred colors and sensitivities to word orderings). For example, gpt-4o-mini often picks the first of two options presented in the prompt regardless of the options' implied likelihood, whereas Llama-3.1-8B picks the second. Our other consistent finding is mode-collapse: Instruction-tuned models often over-allocate probability mass on a single option. These systematic biases introduce non-intuitive model behavior, making models harder for users to understand.
This paper focuses on the development of an advanced intelligent article scoring system that not only assesses the overall quality of written work but also offers detailed feature-based scoring tailored to various article genres. By integrating the pre-trained BERT model with the large language model Chat-GPT, the system gains a deep understanding of both the content and structure of the text, enabling it to provide a thorough evaluation along with targeted suggestions for improvement. Experimental results demonstrate that this system outperforms traditional scoring methods across multiple public datasets, particularly in feature-based assessments, offering a more accurate reflection of the quality of different article types. Moreover, the system generates personalized feedback to assist users in enhancing their writing skills, underscoring the potential and practical value of automated scoring technologies in educational contexts.
Prompt compression condenses contexts while maintaining their informativeness for different usage scenarios. It not only shortens the inference time and reduces computational costs during the usage of large language models, but also lowers expenses when using closed-source models. In a preliminary study, we discover that when instructing language models to compress prompts, different compression styles (e.g., extractive or abstractive) impact performance of compressed prompts on downstream tasks. Building on this insight, we propose Style-Compress, a lightweight framework that adapts a smaller language model to compress prompts for a larger model on a new task without additional training. Our approach iteratively generates and selects effective compressed prompts as task-specific demonstrations through style variation and in-context learning, enabling smaller models to act as efficient compressors with task-specific examples. Style-Compress outperforms two baseline compression models in four tasks: original prompt reconstruction, text summarization, multi-hop QA, and CoT reasoning. In addition, with only 10 samples and 100 queries for adaptation, prompts compressed by Style-Compress achieve performance on par with or better than original prompts at a compression ratio of 0.25 or 0.5.
Electrocardiogram (ECG) interpretation requires specialized expertise, often involving synthesizing insights from ECG signals with complex clinical queries posed in natural language. The scarcity of labeled ECG data coupled with the diverse nature of clinical inquiries presents a significant challenge for developing robust and adaptable ECG diagnostic systems. This work introduces a novel multimodal meta-learning method for few-shot ECG question answering, addressing the challenge of limited labeled data while leveraging the rich knowledge encoded within large language models (LLMs). Our LLM-agnostic approach integrates a pre-trained ECG encoder with a frozen LLM (e.g., LLaMA and Gemma) via a trainable fusion module, enabling the language model to reason about ECG data and generate clinically meaningful answers. Extensive experiments demonstrate superior generalization to unseen diagnostic tasks compared to supervised baselines, achieving notable performance even with limited ECG leads. For instance, in a 5-way 5-shot setting, our method using LLaMA-3.1-8B achieves an accuracy of 84.6%, 77.3%, and 69.6% on single verify, choose and query question types, respectively. These results highlight the potential of our method to enhance clinical ECG interpretation by combining signal processing with the nuanced language understanding capabilities of LLMs, particularly in data-constrained scenarios.
Recently, sharing key-value (KV) cache across layers has been found effective in efficient inference of large language models (LLMs). To systematically investigate different techniques of cross-layer KV sharing, we propose a unified framework that covers several recent methods and their novel variants. We conduct comprehensive experiments on all the configurations of the framework, evaluating their generation throughput and performance in language modeling and downstream tasks. We find that when reducing the size of the KV cache by 2x, most configurations can achieve competitive performance to and higher throughput than standard transformers, but when further reducing the size of the KV cache, pairing queries of all layers with KVs of upper layers can better maintain performance, although it also introduces additional training cost and prefilling latency. We hope that this work will help users choose the appropriate approach according to their requirements and facilitate research on the acceleration of LLM inference.
Neural audio codecs have recently gained popularity because they can represent audio signals with high fidelity at very low bitrates, making it feasible to use language modeling approaches for audio generation and understanding. Residual Vector Quantization (RVQ) has become the standard technique for neural audio compression using a cascade of VQ codebooks. This paper proposes the Multi-Scale Neural Audio Codec, a simple extension of RVQ where the quantizers can operate at different temporal resolutions. By applying a hierarchy of quantizers at variable frame rates, the codec adapts to the audio structure across multiple timescales. This leads to more efficient compression, as demonstrated by extensive objective and subjective evaluations. The code and model weights are open-sourced at https://github.com/hubertsiuzdak/snac.
Spatiotemporal trajectory data is vital for web-of-things services and is extensively collected and analyzed by web-based hardware and platforms. However, issues such as service interruptions and network instability often lead to sparsely recorded trajectories, resulting in a loss of detailed movement data. As a result, recovering these trajectories to restore missing information becomes essential. Despite progress, several challenges remain unresolved. First, the lack of large-scale dense trajectory data hampers the performance of existing deep learning methods, which rely heavily on abundant data for supervised training. Second, current methods struggle to generalize across sparse trajectories with varying sampling intervals, necessitating separate re-training for each interval and increasing computational costs. Third, external factors crucial for the recovery of missing points are not fully incorporated. To address these challenges, we propose a framework called PTR. This framework mitigates the issue of limited dense trajectory data by leveraging the capabilities of pre-trained language models (PLMs). PTR incorporates an explicit trajectory prompt and is trained on datasets with multiple sampling intervals, enabling it to generalize effectively across different intervals in sparse trajectories. To capture external factors, we introduce an implicit trajectory prompt that models road conditions, providing richer information for recovering missing points. Additionally, we present a trajectory embedder that encodes trajectory points and transforms the embeddings of both observed and missing points into a format comprehensible to PLMs. Experimental results on two public trajectory datasets with three sampling intervals demonstrate the efficacy and scalability of PTR.
This study explores the ability of Large Language Model (LLM) agents to detect and correct hallucinations in AI-generated content. A primary agent was tasked with creating a blog about a fictional Danish artist named Flipfloppidy, which was then reviewed by another agent for factual inaccuracies. Most LLMs hallucinated the existence of this artist. Across 4,900 test runs involving various combinations of primary and reviewing agents, advanced AI models such as Llama3-70b and GPT-4 variants demonstrated near-perfect accuracy in identifying hallucinations and successfully revised outputs in 85% to 100% of cases following feedback. These findings underscore the potential of advanced AI models to significantly enhance the accuracy and reliability of generated content, providing a promising approach to improving AI workflow orchestration.
Large Language Models (LLMs) have shown significant potential in designing reward functions for Reinforcement Learning (RL) tasks. However, obtaining high-quality reward code often involves human intervention, numerous LLM queries, or repetitive RL training. To address these issues, we propose CARD, a LLM-driven Reward Design framework that iteratively generates and improves reward function code. Specifically, CARD includes a Coder that generates and verifies the code, while a Evaluator provides dynamic feedback to guide the Coder in improving the code, eliminating the need for human feedback. In addition to process feedback and trajectory feedback, we introduce Trajectory Preference Evaluation (TPE), which evaluates the current reward function based on trajectory preferences. If the code fails the TPE, the Evaluator provides preference feedback, avoiding RL training at every iteration and making the reward function better aligned with the task objective. Empirical results on Meta-World and ManiSkill2 demonstrate that our method achieves an effective balance between task performance and token efficiency, outperforming or matching the baselines across all tasks. On 10 out of 12 tasks, CARD shows better or comparable performance to policies trained with expert-designed rewards, and our method even surpasses the oracle on 3 tasks.
Precise action localization in untrimmed video is vital for fields such as professional sports and minimally invasive surgery, where the delineation of particular motions in recordings can dramatically enhance analysis. But in many cases, large scale datasets with video-label pairs for localization are unavailable, limiting the opportunity to fine-tune video-understanding models. Recent developments in large vision-language models (LVLM) address this need with impressive zero-shot capabilities in a variety of video understanding tasks. However, the adaptation of image-based LVLMs, with their powerful visual question answering capabilities, to action localization in long-form video is still relatively unexplored. To this end, we introduce a true ZEro-shot Action Localization method (ZEAL). Specifically, we leverage the built-in action knowledge of a large language model (LLM) to inflate actions into highly-detailed descriptions of the archetypal start and end of the action. These descriptions serve as queries to LVLM for generating frame-level confidence scores which can be aggregated to produce localization outputs. The simplicity and flexibility of our method lends it amenable to more capable LVLMs as they are developed, and we demonstrate remarkable results in zero-shot action localization on a challenging benchmark, without any training.
Fine-grained video action recognition can be conceptualized as a video-text matching problem. Previous approaches often rely on global video semantics to consolidate video embeddings, which can lead to misalignment in video-text pairs due to a lack of understanding of action semantics at an atomic granularity level. To tackle this challenge, we propose a multi-granularity framework based on two observations: (i) videos with different global semantics may share similar atomic actions or appearances, and (ii) atomic actions within a video can be momentary, slow, or even non-directly related to the global video semantics. Inspired by the concept of storyboarding, which disassembles a script into individual shots, we enhance global video semantics by generating fine-grained descriptions using a pre-trained large language model. These detailed descriptions capture common atomic actions depicted in videos. A filtering metric is proposed to select the descriptions that correspond to the atomic actions present in both the videos and the descriptions. By employing global semantics and fine-grained descriptions, we can identify key frames in videos and utilize them to aggregate embeddings, thereby making the embedding more accurate. Extensive experiments on various video action recognition datasets demonstrate superior performance of our proposed method in supervised, few-shot, and zero-shot settings.
The development of 3D medical vision-language models holds significant potential for disease diagnosis and patient treatment. However, compared to 2D medical images, 3D medical images, such as CT scans, face challenges related to limited training data and high dimension, which severely restrict the progress of 3D medical vision-language models. To address these issues, we collect a large amount of unlabeled 3D CT data and utilize self-supervised learning to construct a 3D visual foundation model for extracting 3D visual features. Then, we apply 3D spatial convolutions to aggregate and project high-level image features, reducing computational complexity while preserving spatial information. We also construct two instruction-tuning datasets based on BIMCV-R and CT-RATE to fine-tune the 3D vision-language model. Our model demonstrates superior performance compared to existing methods in report generation, visual question answering, and disease diagnosis. Code and data will be made publicly available soon.
The growing reproducibility crisis in machine learning has brought forward a need for careful examination of research findings. This paper investigates the claims made by Lei et al. (2023) regarding their proposed method, LICO, for enhancing post-hoc interpretability techniques and improving image classification performance. LICO leverages natural language supervision from a vision-language model to enrich feature representations and guide the learning process. We conduct a comprehensive reproducibility study, employing (Wide) ResNets and established interpretability methods like Grad-CAM and RISE. We were mostly unable to reproduce the authors' results. In particular, we did not find that LICO consistently led to improved classification performance or improvements in quantitative and qualitative measures of interpretability. Thus, our findings highlight the importance of rigorous evaluation and transparent reporting in interpretability research.
Existing tools to detect text generated by a large language model (LLM) have met with certain success, but their performance can drop when dealing with texts in new domains. To tackle this issue, we train a ranking classifier called RoBERTa-Ranker, a modified version of RoBERTa, as a baseline model using a dataset we constructed that includes a wider variety of texts written by humans and generated by various LLMs. We then present a method to fine-tune RoBERTa-Ranker that requires only a small amount of labeled data in a new domain. Experiments show that this fine-tuned domain-aware model outperforms the popular DetectGPT and GPTZero on both in-domain and cross-domain texts, where AI-generated texts may either be in a different domain or generated by a different LLM not used to generate the training datasets. This approach makes it feasible and economical to build a single system to detect AI-generated texts across various domains.