Large language models (LLMs) with one or more fine-tuning phases have become a necessary step to unlock various capabilities, enabling LLMs to follow natural language instructions or align with human preferences. However, it carries the risk of catastrophic forgetting during sequential training, the parametric knowledge or the ability learned in previous stages may be overwhelmed by incoming training data. In this paper, we find that by regularly resetting partial parameters, LLMs can restore some of the original knowledge. Inspired by this, we introduce Half Fine-Tuning (HFT) for LLMs, as a substitute for full fine-tuning (FFT), to mitigate the forgetting issues, where half of the parameters are selected to learn new tasks while the other half are frozen to remain previous knowledge. We provide a feasibility analysis from the perspective of optimization and interpret the parameter selection operation as a regularization term. Without changing the model architecture, HFT could be seamlessly integrated into existing fine-tuning frameworks. Extensive experiments and analysis on supervised fine-tuning, direct preference optimization, and continual learning consistently demonstrate the effectiveness, robustness, and efficiency of HFT. Compared with FFT, HFT not only significantly alleviates the forgetting problem, but also achieves the best performance in a series of downstream benchmarks, with an approximately 30% reduction in training time.

We present COCOLA (Coherence-Oriented Contrastive Learning for Audio), a contrastive learning method for musical audio representations that captures the harmonic and rhythmic coherence between samples. Our method operates at the level of stems (or their combinations) composing music tracks and allows the objective evaluation of compositional models for music in the task of accompaniment generation. We also introduce a new baseline for compositional music generation called CompoNet, based on ControlNet, generalizing the tasks of MSDM, and quantify it against the latter using COCOLA. We release all models trained on public datasets containing separate stems (MUSDB18-HQ, MoisesDB, Slakh2100, and CocoChorales).

Vision and Language Models (VLMs) continue to demonstrate remarkable zero-shot (ZS) performance across various tasks. However, many probing studies have revealed that even the best-performing VLMs struggle to capture aspects of compositional scene understanding, lacking the ability to properly ground and localize linguistic phrases in images. Recent VLM advancements include scaling up both model and dataset sizes, additional training objectives and levels of supervision, and variations in the model architectures. To characterize the grounding ability of VLMs, such as phrase grounding, referring expressions comprehension, and relationship understanding, Pointing Game has been used as an evaluation metric for datasets with bounding box annotations. In this paper, we introduce a novel suite of quantitative metrics that utilize GradCAM activations to rigorously evaluate the grounding capabilities of pre-trained VLMs like CLIP, BLIP, and ALBEF. These metrics offer an explainable and quantifiable approach for a more detailed comparison of the zero-shot capabilities of VLMs and enable measuring models' grounding uncertainty. This characterization reveals interesting tradeoffs between the size of the model, the dataset size, and their performance.

This work introduces LAB (Large-scale Alignment for chatBots), a novel methodology designed to overcome the scalability challenges in the instruction-tuning phase of large language model (LLM) training. Leveraging a taxonomy-guided synthetic data generation process and a multi-phase tuning framework, LAB significantly reduces reliance on expensive human annotations and proprietary models like GPT-4. We demonstrate that LAB-trained models can achieve competitive performance across several benchmarks compared to models trained with traditional human-annotated or GPT-4 generated synthetic data. Large language models (LLMs) have achieved remarkable levels of success in various natural language processing (NLP) applications, including question-answering, entity extraction, and summarization . This has been made possible, in large part, by the introduction of the transformer architecture, which can leverage large amounts of unlabeled, unstructured data, enabling the scaling of LLMs to billions, or even trillions of parameters. LLMs are typically trained in phases: a self-supervised pre-training phase, followed by supervised alignment tuning phases.

Explanatory inference is the creation and evaluation of hypotheses that provide explanations, and is sometimes known as abduction or abductive inference. Generative AI is a new set of artificial intelligence models based on novel algorithms for generating text, images, and sounds. This paper proposes a set of benchmarks for assessing the ability of AI programs to perform explanatory inference, and uses them to determine the extent to which ChatGPT, a leading generative AI model, is capable of making explanatory inferences. Tests on the benchmarks reveal that ChatGPT performs creative and evaluative inferences in many domains, although it is limited to verbal and visual modalities. Claims that ChatGPT and similar models are incapable of explanation, understanding, causal reasoning, meaning, and creativity are rebutted.

The integration of Large Language Models (LLMs) into various software applications raises concerns about their potential biases. Typically, those models are trained on a vast amount of data scrapped from forums, websites, social media and other internet sources, which may instill harmful and discriminating behavior into the model. To address this issue, we present LangBiTe, a testing platform to systematically assess the presence of biases within an LLM. LangBiTe enables development teams to tailor their test scenarios, and automatically generate and execute the test cases according to a set of user-defined ethical requirements. Each test consists of a prompt fed into the LLM and a corresponding test oracle that scrutinizes the LLM's response for the identification of biases. LangBite provides users with the bias evaluation of LLMs, and end-to-end traceability between the initial ethical requirements and the insights obtained.

Beyond scaling base models with more data or parameters, fine-tuned adapters provide an alternative way to generate high fidelity, custom images at reduced costs. As such, adapters have been widely adopted by open-source communities, accumulating a database of over 100K adapters-most of which are highly customized with insufficient descriptions. This paper explores the problem of matching the prompt to a set of relevant adapters, built on recent work that highlight the performance gains of composing adapters. We introduce Stylus, which efficiently selects and automatically composes task-specific adapters based on a prompt's keywords. Stylus outlines a three-stage approach that first summarizes adapters with improved descriptions and embeddings, retrieves relevant adapters, and then further assembles adapters based on prompts' keywords by checking how well they fit the prompt. To evaluate Stylus, we developed StylusDocs, a curated dataset featuring 75K adapters with pre-computed adapter embeddings. In our evaluation on popular Stable Diffusion checkpoints, Stylus achieves greater CLIP-FID Pareto efficiency and is twice as preferred, with humans and multimodal models as evaluators, over the base model. See stylus-diffusion.github.io for more.

The networking field is characterized by its high complexity and rapid iteration, requiring extensive expertise to accomplish network tasks, ranging from network design, configuration, diagnosis and security. The inherent complexity of these tasks, coupled with the ever-changing landscape of networking technologies and protocols, poses significant hurdles for traditional machine learning-based methods. These methods often struggle to generalize and automate complex tasks in networking, as they require extensive labeled data, domain-specific feature engineering, and frequent retraining to adapt to new scenarios. However, the recent emergence of large language models (LLMs) has sparked a new wave of possibilities in addressing these challenges. LLMs have demonstrated remarkable capabilities in natural language understanding, generation, and reasoning. These models, trained on extensive data, can benefit the networking domain. Some efforts have already explored the application of LLMs in the networking domain and revealed promising results. By reviewing recent advances, we present an abstract workflow to describe the fundamental process involved in applying LLM for Networking. We introduce the highlights of existing works by category and explain in detail how they operate at different stages of the workflow. Furthermore, we delve into the challenges encountered, discuss potential solutions, and outline future research prospects. We hope that this survey will provide insight for researchers and practitioners, promoting the development of this interdisciplinary research field.

While language models (LMs) offer significant capability in zero-shot reasoning tasks across a wide range of domains, they do not perform satisfactorily in problems which requires multi-step reasoning. Previous approaches to mitigate this involves breaking a larger, multi-step task into sub-tasks and asking the language model to generate proposals ("thoughts") for each sub-task and using exhaustive planning approaches such as DFS to compose a solution. In this work, we leverage this idea to introduce two new contributions: first, we formalize a planning-based approach to perform multi-step problem solving with LMs via Partially Observable Markov Decision Processes (POMDPs), with the LM's own reflections about the value of a state used as a search heuristic; second, leveraging the online POMDP solver POMCP, we demonstrate a superior success rate of 89.4% on the Game of 24 task as compared to existing approaches while also offering better anytime performance characteristics than fixed tree-search which is used previously. Taken together, these contributions allow modern LMs to decompose and solve larger-scale reasoning tasks more effectively.

Generative large-scale language models create the fifth paradigm of scientific research, organically combine data science and computational intelligence, transform the research paradigm of natural language processing and multimodal information processing, promote the new trend of AI-enabled social science research, and provide new ideas for digital humanities research and application. This article profoundly explores the application of large-scale language models in digital humanities research, revealing their significant potential in ancient book protection, intelligent processing, and academic innovation. The article first outlines the importance of ancient book resources and the necessity of digital preservation, followed by a detailed introduction to developing large-scale language models, such as ChatGPT, and their applications in document management, content understanding, and cross-cultural research. Through specific cases, the article demonstrates how AI can assist in the organization, classification, and content generation of ancient books. Then, it explores the prospects of AI applications in artistic innovation and cultural heritage preservation. Finally, the article explores the challenges and opportunities in the interaction of technology, information, and society in the digital humanities triggered by AI technologies.