Whisper and other large-scale automatic speech recognition models have made significant progress in performance. However, their performance on many low-resource languages, such as Kazakh, is not satisfactory. It is worth researching how to utilize low-cost data to improve the performance of Whisper on under-represented languages. In this study, we utilized easily accessible unpaired speech and text data and combined the language model GPT with Whisper on Kazakh. We implemented end of transcript (EOT) judgment modification and hallucination penalty to improve the performance of speech recognition. Further, we employed the decoding average token log probability as a criterion to select samples from unlabeled speech data and used pseudo-labeled data to fine-tune the model to further improve its performance. Ultimately, we achieved more than 10\% absolute WER reduction in multiple experiments, and the whole process has the potential to be generalized to other under-represented languages.
Unified graph representation learning aims to produce node embeddings, which can be applied to multiple downstream applications. However, existing studies based on graph neural networks and language models either suffer from the limitations of numerous training needed toward specific downstream predictions or have shallow semantic features. In this work, we propose a novel Path-LLM model to learn unified graph representation, which leverages a powerful large language model (LLM) to incorporate our proposed path features. Our Path-LLM framework consists of several well-designed techniques. First, we develop a new mechanism of long-to-short shortest path (L2SP) selection, which covers essential connections between different dense groups. An in-depth comparison of different path selection plans is offered to illustrate the strength of our designed L2SP. Then, we design path textualization to obtain L2SP-based training texts. Next, we feed the texts into a self-supervised LLM training process to learn embeddings. Extensive experiments on benchmarks validate the superiority of Path-LLM against the state-of-the-art WalkLM method on two classical graph learning tasks (node classification and link prediction) and one NP-hard graph query processing task (keyword search), meanwhile saving more than 90% of training paths.
The first edition of the workshop on Large Language Model for Evaluation in Information Retrieval (LLM4Eval 2024) took place in July 2024, co-located with the ACM SIGIR Conference 2024 in the USA (SIGIR 2024). The aim was to bring information retrieval researchers together around the topic of LLMs for evaluation in information retrieval that gathered attention with the advancement of large language models and generative AI. Given the novelty of the topic, the workshop was focused around multi-sided discussions, namely panels and poster sessions of the accepted proceedings papers.
Large language models (LLMs) have demonstrated prowess in a wide range of tasks. However, many LLMs exhibit significant performance discrepancies between high- and low-resource languages. To mitigate this challenge, we present FuxiTranyu, an open-source multilingual LLM, which is designed to satisfy the need of the research community for balanced and high-performing multilingual capabilities. The base model, FuxiTranyu-8B, features 8 billion parameters and is trained from scratch on meticulously balanced multilingual data that contains 600 billion tokens covering 43 natural languages and 16 programming languages. We also develop two instruction-tuned models: FuxiTranyu-8B-SFT which is fine-tuned on a diverse multilingual instruction dataset, and FuxiTranyu-8B-DPO which is further refined with DPO on a preference dataset for enhanced alignment ability. Extensive experiments on a wide range of multilingual benchmarks demonstrate the competitive performance of FuxiTranyu against existing multilingual LLMs, e.g., BLOOM-7B, PolyLM-13B, and Mistral-7B-Instruct. Both neuron and representation interpretability analyses reveal that FuxiTranyu achieves consistent multilingual representations across languages. To promote further research into multilingual LLMs, we release both the base and instruction-tuned FuxiTranyu models together with 58 pre-training checkpoints at HuggingFace (see https://huggingface.co/TJUNLP/FuxiTranyu-8B) and Github (see https://github.com/tjunlp-lab/FuxiTranyu).
In this study, we generate and maintain a database of 10 million virtual lipids through METiS's in-house de novo lipid generation algorithms and lipid virtual screening techniques. These virtual lipids serve as a corpus for pre-training, lipid representation learning, and downstream task knowledge transfer, culminating in state-of-the-art LNP property prediction performance. We propose LipidBERT, a BERT-like model pre-trained with the Masked Language Model (MLM) and various secondary tasks. Additionally, we compare the performance of embeddings generated by LipidBERT and PhatGPT, our GPT-like lipid generation model, on downstream tasks. The proposed bilingual LipidBERT model operates in two languages: the language of ionizable lipid pre-training, using in-house dry-lab lipid structures, and the language of LNP fine-tuning, utilizing in-house LNP wet-lab data. This dual capability positions LipidBERT as a key AI-based filter for future screening tasks, including new versions of METiS de novo lipid libraries and, more importantly, candidates for in vivo testing for orgran-targeting LNPs. To the best of our knowledge, this is the first successful demonstration of the capability of a pre-trained language model on virtual lipids and its effectiveness in downstream tasks using web-lab data. This work showcases the clever utilization of METiS's in-house de novo lipid library as well as the power of dry-wet lab integration.
As large language model (LLM) inference continues to demand increasing computational resources, there is a rapidly growing trend toward using low-bit weights to reduce memory footprint and improve inference efficiency. However, low-bit LLMs introduce the need for mixed-precision general matrix multiplication (mpGEMM), which involves multiplying low-precision weights with higher-precision activations - a critical yet under-explored operation. Current hardware lacks native support for mpGEMM, leading to inefficient dequantization-based implementations. To address this, we explore a lookup table (LUT)-based approach to accelerate mpGEMM. While conventional LUT implementations fall short in performance and flexibility, we propose LUT Tensor Core, a software-hardware co-designed solution optimized for low-bit LLM inference. On the software side, we introduce operator fusion and table symmetrization techniques to optimize LUT generation and storage. On the hardware side, LUT Tensor Core adopts an elongated tiling shape to maximize table reuse and employs a bit-serial architecture to flexibly support a variety of precision combinations. Additionally, we design an end-to-end compilation stack with custom instructions to enable efficient code generation and optimization for LUT-based mpGEMM. Experimental results on low-bit LLMs such as BitNet and LLaMA demonstrate that LUT Tensor Core delivers over an order-of-magnitude improvement in both compute density and energy efficiency.
Existing sample-based methods, like influence functions and representer points, measure the importance of a training point by approximating the effect of its removal from training. As such, they are skewed towards outliers and points that are very close to the decision boundaries. The explanations provided by these methods are often static and not specific enough for different test points. In this paper, we propose a method to generate an explanation in the form of support spectrums which are based on two main ideas: the support sets and a global-to-local importance measure. The support set is the set of training points, in the predicted class, that ``lie in between'' the test point and training points in the other classes. They indicate how well the test point can be distinguished from the points not in the predicted class. The global-to-local importance measure is obtained by decoupling existing methods into the global and local components which are then used to select the points in the support set. Using this method, we are able to generate explanations that are tailored to specific test points. In the experiments, we show the effectiveness of the method in image classification and text generation tasks.
Speculative decoding has emerged as a widely adopted method to accelerate large language model inference without sacrificing the quality of the model outputs. While this technique has facilitated notable speed improvements by enabling parallel sequence verification, its efficiency remains inherently limited by the reliance on incremental token generation in existing draft models. To overcome this limitation, this paper proposes an adaptation of speculative decoding which uses discrete diffusion models to generate draft sequences. This allows parallelization of both the drafting and verification steps, providing significant speed-ups to the inference process. Our proposed approach, Speculative Diffusion Decoding (SpecDiff), is validated on standard language generation benchmarks and empirically demonstrated to provide a up to 8.7x speed-up over standard generation processes and up to 2.5x speed-up over existing speculative decoding approaches.
In this paper, we introduce PhishLang, the first fully client-side anti-phishing framework built on a lightweight ensemble framework that utilizes advanced language models to analyze the contextual features of a website's source code and URL. Unlike traditional heuristic or machine learning approaches that rely on static features and struggle to adapt to evolving threats, or deep learning models that are computationally intensive, our approach utilizes MobileBERT, a fast and memory-efficient variant of the BERT architecture, to capture nuanced features indicative of phishing attacks. To further enhance detection accuracy, PhishLang employs a multi-modal ensemble approach, combining both the URL and Source detection models. This architecture ensures robustness by allowing one model to compensate for scenarios where the other may fail, or if both models provide ambiguous inferences. As a result, PhishLang excels at detecting both regular and evasive phishing threats, including zero-day attacks, outperforming popular anti-phishing tools, while operating without relying on external blocklists and safeguarding user privacy by ensuring that browser history remains entirely local and unshared. We release PhishLang as a Chromium browser extension and also open-source the framework to aid the research community.
Recently, there has been an extensive research effort in building efficient large language model (LLM) inference serving systems. These efforts not only include innovations in the algorithm and software domains but also constitute developments of various hardware acceleration techniques. Nevertheless, there is a lack of simulation infrastructure capable of accurately modeling versatile hardware-software behaviors in LLM serving systems without extensively extending the simulation time. This paper aims to develop an effective simulation tool, called LLMServingSim, to support future research in LLM serving systems. In designing LLMServingSim, we focus on two limitations of existing simulators: (1) they lack consideration of the dynamic workload variations of LLM inference serving due to its autoregressive nature, and (2) they incur repetitive simulations without leveraging algorithmic redundancies in LLMs. To address these limitations, LLMServingSim simulates the LLM serving in the granularity of iterations, leveraging the computation redundancies across decoder blocks and reusing the simulation results from previous iterations. Additionally, LLMServingSim provides a flexible framework that allows users to plug in any accelerator compiler-and-simulation stacks for exploring various system designs with heterogeneous processors. Our experiments demonstrate that LLMServingSim produces simulation results closely following the performance behaviors of real GPU-based LLM serving system with less than 14.7% error rate, while offering 91.5x faster simulation speed compared to existing accelerator simulators.
Foundation Models (FMs), e.g., large language models, possess attributes of intelligence which offer promise to endow a robot with the contextual understanding necessary to navigate complex, unstructured tasks in the wild. We see three core challenges in the future of space robotics that motivate building an FM for the space robotics community: 1) Scalability of ground-in-the-loop operations; 2) Generalizing prior knowledge to novel environments; and 3) Multi-modality in tasks and sensor data. As a first-step towards a space foundation model, we programmatically augment three extraterrestrial databases with fine-grained language annotations inspired by the sensory reasoning necessary to e.g., identify a site of scientific interest on Mars, building a synthetic dataset of visual-question-answer and visual instruction-following tuples. We fine-tune a pre-trained LLaVA 13B checkpoint on our augmented dataset to adapt a Vision-Language Model (VLM) to the visual semantic features in an extraterrestrial environment, demonstrating FMs as a tool for specialization and enhancing a VLM's zero-shot performance on unseen task types in comparison to state-of-the-art VLMs. Ablation studies show that fine-tuning the language backbone and vision-language adapter in concert is key to facilitate adaption while a small percentage, e.g., 20%, of the pre-training data can be used to safeguard against catastrophic forgetting.
One of the grand challenges of artificial general intelligence is developing agents capable of conducting scientific research and discovering new knowledge. While frontier models have already been used as aides to human scientists, e.g. for brainstorming ideas, writing code, or prediction tasks, they still conduct only a small part of the scientific process. This paper presents the first comprehensive framework for fully automatic scientific discovery, enabling frontier large language models to perform research independently and communicate their findings. We introduce The AI Scientist, which generates novel research ideas, writes code, executes experiments, visualizes results, describes its findings by writing a full scientific paper, and then runs a simulated review process for evaluation. In principle, this process can be repeated to iteratively develop ideas in an open-ended fashion, acting like the human scientific community. We demonstrate its versatility by applying it to three distinct subfields of machine learning: diffusion modeling, transformer-based language modeling, and learning dynamics. Each idea is implemented and developed into a full paper at a cost of less than $15 per paper. To evaluate the generated papers, we design and validate an automated reviewer, which we show achieves near-human performance in evaluating paper scores. The AI Scientist can produce papers that exceed the acceptance threshold at a top machine learning conference as judged by our automated reviewer. This approach signifies the beginning of a new era in scientific discovery in machine learning: bringing the transformative benefits of AI agents to the entire research process of AI itself, and taking us closer to a world where endless affordable creativity and innovation can be unleashed on the world's most challenging problems. Our code is open-sourced at https://github.com/SakanaAI/AI-Scientist
Representation Misdirection for Unlearning (RMU), which steers model representation in the intermediate layer to a target random representation, is an effective method for large language model (LLM) unlearning. Despite its high performance, the underlying cause and explanation remain underexplored. In this paper, we theoretically demonstrate that steering forget representations in the intermediate layer reduces token confidence, causing LLMs to generate wrong or nonsense responses. We investigate how the coefficient influences the alignment of forget-sample representations with the random direction and hint at the optimal coefficient values for effective unlearning across different network layers. We show that RMU unlearned models are robust against adversarial jailbreak attacks. Furthermore, our empirical analysis shows that RMU is less effective when applied to the middle and later layers in LLMs. To resolve this drawback, we propose Adaptive RMU -- a simple yet effective alternative method that makes unlearning effective with most layers. Extensive experiments demonstrate that Adaptive RMU significantly improves the unlearning performance compared to prior art while incurring no additional computational cost.
We propose a novel approach for decision making problems leveraging the generalization capabilities of large language models (LLMs). Traditional methods such as expert systems, planning algorithms, and reinforcement learning often exhibit limited generalization, typically requiring the training of new models for each unique task. In contrast, LLMs demonstrate remarkable success in generalizing across varied language tasks, inspiring a new strategy for training decision making models. Our approach, referred to as "Learning then Using" (LTU), entails a two-stage process. Initially, the \textit{learning} phase develops a robust foundational decision making model by integrating diverse knowledge from various domains and decision making contexts. The subsequent \textit{using} phase refines this foundation model for specific decision making scenarios. Distinct from other studies that employ LLMs for decision making through supervised learning, our LTU method embraces a versatile training methodology that combines broad pre-training with targeted fine-tuning. Experiments in e-commerce domains such as advertising and search optimization have shown that LTU approach outperforms traditional supervised learning regimes in decision making capabilities and generalization. The LTU approach is the first practical training architecture for both single-step and multi-step decision making tasks combined with LLMs, which can be applied beyond game and robot domains. It provides a robust and adaptable framework for decision making, enhances the effectiveness and flexibility of various systems in tackling various challenges.
Guitar tablatures enrich the structure of traditional music notation by assigning each note to a string and fret of a guitar in a particular tuning, indicating precisely where to play the note on the instrument. The problem of generating tablature from a symbolic music representation involves inferring this string and fret assignment per note across an entire composition or performance. On the guitar, multiple string-fret assignments are possible for most pitches, which leads to a large combinatorial space that prevents exhaustive search approaches. Most modern methods use constraint-based dynamic programming to minimize some cost function (e.g.\ hand position movement). In this work, we introduce a novel deep learning solution to symbolic guitar tablature estimation. We train an encoder-decoder Transformer model in a masked language modeling paradigm to assign notes to strings. The model is first pre-trained on DadaGP, a dataset of over 25K tablatures, and then fine-tuned on a curated set of professionally transcribed guitar performances. Given the subjective nature of assessing tablature quality, we conduct a user study amongst guitarists, wherein we ask participants to rate the playability of multiple versions of tablature for the same four-bar excerpt. The results indicate our system significantly outperforms competing algorithms.
Performance of NLP systems is typically evaluated by collecting a large-scale dataset by means of crowd-sourcing to train a data-driven model and evaluate it on a held-out portion of the data. This approach has been shown to suffer from spurious correlations and the lack of challenging examples that represent the diversity of natural language. Instead, we examine a framework for evaluating optimised models in training-set free setting on synthetically generated challenge sets. We find that despite the simplicity of the generation method, the data can compete with crowd-sourced datasets with regard to naturalness and lexical diversity for the purpose of evaluating the linguistic capabilities of MRC models. We conduct further experiments and show that state-of-the-art language model-based MRC systems can learn to succeed on the challenge set correctly, although, without capturing the general notion of the evaluated phenomenon.
Episodic structures are inherently interpretable and adaptable to evolving large-scale key events. However, state-of-the-art automatic event detection methods overlook event episodes and, therefore, struggle with these crucial characteristics. This paper introduces a novel task, episode detection, aimed at identifying episodes from a news corpus containing key event articles. An episode describes a cohesive cluster of core entities (e.g., "protesters", "police") performing actions at a specific time and location. Furthermore, an episode is a significant part of a larger group of episodes under a particular key event. Automatically detecting episodes is challenging because, unlike key events and atomic actions, we cannot rely on explicit mentions of times and locations to distinguish between episodes or use semantic similarity to merge inconsistent episode co-references. To address these challenges, we introduce EpiMine, an unsupervised episode detection framework that (1) automatically identifies the most salient, key-event-relevant terms and segments, (2) determines candidate episodes in an article based on natural episodic partitions estimated through shifts in discriminative term combinations, and (3) refines and forms final episode clusters using large language model-based reasoning on the candidate episodes. We construct three diverse, real-world event datasets annotated at the episode level. EpiMine outperforms all baselines on these datasets by an average 59.2% increase across all metrics.
Large Language Models (LLMs) offer significant potential for clinical symptom extraction, but their deployment in healthcare settings is constrained by privacy concerns, computational limitations, and operational costs. This study investigates the optimization of compact LLMs for cancer toxicity symptom extraction using a novel iterative refinement approach. We employ a student-teacher architecture, utilizing Zephyr-7b-beta and Phi3-mini-128 as student models and GPT-4o as the teacher, to dynamically select between prompt refinement, Retrieval-Augmented Generation (RAG), and fine-tuning strategies. Our experiments on 294 clinical notes covering 12 post-radiotherapy toxicity symptoms demonstrate the effectiveness of this approach. The RAG method proved most efficient, improving average accuracy scores from 0.32 to 0.73 for Zephyr-7b-beta and from 0.40 to 0.87 for Phi3-mini-128 during refinement. In the test set, both models showed an approximate 0.20 increase in accuracy across symptoms. Notably, this improvement was achieved at a cost 45 times lower than GPT-4o for Zephyr and 79 times lower for Phi-3. These results highlight the potential of iterative refinement techniques in enhancing the capabilities of compact LLMs for clinical applications, offering a balance between performance, cost-effectiveness, and privacy preservation in healthcare settings.
This study utilizes the advanced capabilities of the GPT-4 multimodal Large Language Model (LLM) to explore its potential in iris recognition - a field less common and more specialized than face recognition. By focusing on this niche yet crucial area, we investigate how well AI tools like ChatGPT can understand and analyze iris images. Through a series of meticulously designed experiments employing a zero-shot learning approach, the capabilities of ChatGPT-4 was assessed across various challenging conditions including diverse datasets, presentation attacks, occlusions such as glasses, and other real-world variations. The findings convey ChatGPT-4's remarkable adaptability and precision, revealing its proficiency in identifying distinctive iris features, while also detecting subtle effects like makeup on iris recognition. A comparative analysis with Gemini Advanced - Google's AI model - highlighted ChatGPT-4's better performance and user experience in complex iris analysis tasks. This research not only validates the use of LLMs for specialized biometric applications but also emphasizes the importance of nuanced query framing and interaction design in extracting significant insights from biometric data. Our findings suggest a promising path for future research and the development of more adaptable, efficient, robust and interactive biometric security solutions.
Significant research efforts have been made to scale and improve vision-language model (VLM) training approaches. Yet, with an ever-growing number of benchmarks, researchers are tasked with the heavy burden of implementing each protocol, bearing a non-trivial computational cost, and making sense of how all these benchmarks translate into meaningful axes of progress. To facilitate a systematic evaluation of VLM progress, we introduce UniBench: a unified implementation of 50+ VLM benchmarks spanning a comprehensive range of carefully categorized capabilities from object recognition to spatial awareness, counting, and much more. We showcase the utility of UniBench for measuring progress by evaluating nearly 60 publicly available vision-language models, trained on scales of up to 12.8B samples. We find that while scaling training data or model size can boost many vision-language model capabilities, scaling offers little benefit for reasoning or relations. Surprisingly, we also discover today's best VLMs struggle on simple digit recognition and counting tasks, e.g. MNIST, which much simpler networks can solve. Where scale falls short, we find that more precise interventions, such as data quality or tailored-learning objectives offer more promise. For practitioners, we also offer guidance on selecting a suitable VLM for a given application. Finally, we release an easy-to-run UniBench code-base with the full set of 50+ benchmarks and comparisons across 59 models as well as a distilled, representative set of benchmarks that runs in 5 minutes on a single GPU.