Question answering (QA) has become an important application in the advanced development of large language models. General pre-trained large language models for question-answering are not trained to properly understand the knowledge or terminology for a specific domain, such as finance, healthcare, education, and customer service for a product. To better cater to domain-specific understanding, we build an in-house question-answering system for Adobe products. We propose a novel framework to compile a large question-answer database and develop the approach for retrieval-aware finetuning of a Large Language model. We showcase that fine-tuning the retriever leads to major improvements in the final generation. Our overall approach reduces hallucinations during generation while keeping in context the latest retrieval information for contextual grounding.

Cascades and speculative decoding are two common approaches to improving language models' inference efficiency. Both approaches involve interleaving models of different sizes, but via fundamentally distinct mechanisms: cascades employ a deferral rule that invokes the larger model only for "hard" inputs, while speculative decoding uses speculative execution to primarily invoke the larger model in parallel verification mode. These mechanisms offer different benefits: empirically, cascades are often capable of yielding better quality than even the larger model, while theoretically, speculative decoding offers a guarantee of quality-neutrality. In this paper, we leverage the best of both these approaches by designing new speculative cascading techniques that implement their deferral rule through speculative execution. We characterize the optimal deferral rule for our speculative cascades, and employ a plug-in approximation to the optimal rule. Through experiments with T5 models on benchmark language tasks, we show that the proposed approach yields better cost-quality trade-offs than cascading and speculative decoding baselines.

The burgeoning size of Large Language Models (LLMs) has led to enhanced capabilities in generating responses, albeit at the expense of increased inference times and elevated resource demands. Existing methods of acceleration, predominantly hinged on knowledge distillation, generally necessitate fine-tuning of considerably large models, such as Llama-7B, posing a challenge for average users. Furthermore, present techniques for expediting inference and reducing costs operate independently. To address these issues, we introduce a novel and intuitive Guidance-based Knowledge Transfer (GKT) framework. This approach leverages a larger LLM as a ''teacher'' to create guidance prompts, paired with a smaller ''student'' model to finalize responses. Remarkably, GKT requires no fine-tuning and doesn't necessitate the teacher and student models to have the same vocabulary, allowing for extensive batch generation to accelerate the process while ensuring user customization. GKT can be seamlessly integrated into cloud-edge collaboration architectures, and is versatile enough for plug-and-play application across various models. It excels in both efficiency and affordability, epitomizing a ''cheap and cheerful'' solution. GKT achieves a maximum accuracy improvement of 14.18%, along with a 10.72 times speed-up on GSM8K and an accuracy improvement of 14.00 % along with a 7.73 times speed-up in CSQA. When utilizing ChatGPT as teacher model and Llama2-70B as the student model, we can achieve 95.00% of ChatGPT's performance at 52% of the cost. The results highlight substantial enhancements in accuracy and processing speed on the GSM8K and CSQA datasets, surpassing the performance of using either the student or teacher models in isolation.

In this paper, we make a contribution that can be understood from two perspectives: from an NLP perspective, we introduce a small challenge dataset for NLI with large lexical overlap, which minimises the possibility of models discerning entailment solely based on token distinctions, and show that GPT-4 and Llama 2 fail it with strong bias. We then create further challenging sub-tasks in an effort to explain this failure. From a Computational Linguistics perspective, we identify a group of constructions with three classes of adjectives which cannot be distinguished by surface features. This enables us to probe for LLM's understanding of these constructions in various ways, and we find that they fail in a variety of ways to distinguish between them, suggesting that they don't adequately represent their meaning or capture the lexical properties of phrasal heads.

We present a generative dialogue system capable of operating in a full-duplex manner, allowing for seamless interaction. It is based on a large language model (LLM) carefully aligned to be aware of a perception module, a motor function module, and the concept of a simple finite state machine (called neural FSM) with two states. The perception and motor function modules operate simultaneously, allowing the system to simultaneously speak and listen to the user. The LLM generates textual tokens for inquiry responses and makes autonomous decisions to start responding to, wait for, or interrupt the user by emitting control tokens to the neural FSM. All these tasks of the LLM are carried out as next token prediction on a serialized view of the dialogue in real-time. In automatic quality evaluations simulating real-life interaction, the proposed system reduces the average conversation response latency by more than 3 folds compared with LLM-based half-duplex dialogue systems while responding within less than 500 milliseconds in more than 50% of evaluated interactions. Running a LLM with only 8 billion parameters, our system exhibits a 8% higher interruption precision rate than the best available commercial LLM for voice-based dialogue.

The ability of Large Language Models (LLMs) to process and generate coherent text is markedly weakened when the number of input tokens exceeds their pretraining length. Given the expensive overhead of finetuning large-scale models with longer sequences, we propose Dual Chunk Attention (DCA), which enables Llama2 70B to support context windows of more than 100k tokens without continual training. By decomposing the attention computation for long sequences into chunk-based modules, DCA manages to effectively capture the relative positional information of tokens within the same chunk (Intra-Chunk) and across distinct chunks (Inter-Chunk), as well as integrates seamlessly with Flash Attention. In addition to its impressive extrapolation capability, DCA achieves performance on practical long-context tasks that is comparable to or even better than that of finetuned models. When compared with proprietary models, our training-free 70B model attains 94% of the performance of gpt-3.5-16k, indicating it is a viable open-source alternative. All code and data used in this work are released at \url{https://github.com/HKUNLP/ChunkLlama}.

We present STAT: a simple algorithm to prune transformer models without any fine-tuning. STAT eliminates both attention heads and neurons from the network, while preserving accuracy by calculating a correction to the weights of the next layer. Each layer block in the network is compressed using a series of principled matrix factorizations that preserve the network structure. Our entire algorithm takes minutes to compress BERT, and less than three hours to compress models with 7B parameters using a single GPU. Using only several hundred data examples, STAT preserves the output of the network and improves upon existing gradient-free pruning methods. It is even competitive with methods that include significant fine-tuning. We demonstrate our method on both encoder and decoder architectures, including BERT, DistilBERT, and Llama-2 using benchmarks such as GLUE, Squad, WikiText2.

Large Language Models (LLMs) have exhibited remarkable proficiency in generating code. However, the misuse of LLM-generated (Synthetic) code has prompted concerns within both educational and industrial domains, highlighting the imperative need for the development of synthetic code detectors. Existing methods for detecting LLM-generated content are primarily tailored for general text and often struggle with code content due to the distinct grammatical structure of programming languages and massive "low-entropy" tokens. Building upon this, our work proposes a novel zero-shot synthetic code detector based on the similarity between the code and its rewritten variants. Our method relies on the intuition that the differences between the LLM-rewritten and original codes tend to be smaller when the original code is synthetic. We utilize self-supervised contrastive learning to train a code similarity model and assess our approach on two synthetic code detection benchmarks. Our results demonstrate a notable enhancement over existing synthetic content detectors designed for general texts, with an improvement of 20.5% in the APPS benchmark and 29.1% in the MBPP benchmark.

Recent advances in text-to-music editing, which employ text queries to modify music (e.g. by changing its style or adjusting instrumental components), present unique challenges and opportunities for AI-assisted music creation. Previous approaches in this domain have been constrained by the necessity to train specific editing models from scratch, which is both resource-intensive and inefficient; other research uses large language models to predict edited music, resulting in imprecise audio reconstruction. To Combine the strengths and address these limitations, we introduce Instruct-MusicGen, a novel approach that finetunes a pretrained MusicGen model to efficiently follow editing instructions such as adding, removing, or separating stems. Our approach involves a modification of the original MusicGen architecture by incorporating a text fusion module and an audio fusion module, which allow the model to process instruction texts and audio inputs concurrently and yield the desired edited music. Remarkably, Instruct-MusicGen only introduces 8% new parameters to the original MusicGen model and only trains for 5K steps, yet it achieves superior performance across all tasks compared to existing baselines, and demonstrates performance comparable to the models trained for specific tasks. This advancement not only enhances the efficiency of text-to-music editing but also broadens the applicability of music language models in dynamic music production environments.

Growing interest and investment in the capabilities of foundation models has positioned such systems to impact a wide array of public services. Alongside these opportunities is the risk that these systems reify existing power imbalances and cause disproportionate harm to marginalized communities. Participatory approaches hold promise to instead lend agency and decision-making power to marginalized stakeholders. But existing approaches in participatory AI/ML are typically deeply grounded in context - how do we apply these approaches to foundation models, which are, by design, disconnected from context? Our paper interrogates this question. First, we examine existing attempts at incorporating participation into foundation models. We highlight the tension between participation and scale, demonstrating that it is intractable for impacted communities to meaningfully shape a foundation model that is intended to be universally applicable. In response, we develop a blueprint for participatory foundation models that identifies more local, application-oriented opportunities for meaningful participation. In addition to the "foundation" layer, our framework proposes the "subfloor'' layer, in which stakeholders develop shared technical infrastructure, norms and governance for a grounded domain, and the "surface'' layer, in which affected communities shape the use of a foundation model for a specific downstream task. The intermediate "subfloor'' layer scopes the range of potential harms to consider, and affords communities more concrete avenues for deliberation and intervention. At the same time, it avoids duplicative effort by scaling input across relevant use cases. Through three case studies in clinical care, financial services, and journalism, we illustrate how this multi-layer model can create more meaningful opportunities for participation than solely intervening at the foundation layer.