Unlocking Efficiency in Large Language Model Inference: A Comprehensive Survey of Speculative Decoding

To mitigate the high inference latency stemming from autoregressive decoding in Large Language Models (LLMs), Speculative Decoding has emerged as a novel decoding paradigm for LLM inference. In each decoding step, this method first efficiently drafts several future tokens and then verifies them in parallel. Unlike autoregressive decoding, Speculative Decoding facilitates the simultaneous decoding of multiple tokens per step, thereby accelerating inference. This paper presents a comprehensive overview and analysis of this promising decoding paradigm. We begin by providing a formal definition and formulation of Speculative Decoding. Then, we organize in-depth discussions on its key facets, including current leading techniques, the challenges faced, and potential future directions in this field. We aim for this work to serve as a catalyst for further research on Speculative Decoding, ultimately contributing to more efficient LLM inference.

ALYMPICS: Language Agents Meet Game Theory

This paper introduces Alympics, a platform that leverages Large Language Model (LLM) agents to facilitate investigations in game theory. By employing LLMs and autonomous agents to simulate human behavior and enable multi-agent collaborations, we can construct realistic and dynamic models of human interactions for game theory hypothesis formulating and testing. To demonstrate this, we present and implement a survival game involving unequal competition for limited resources. Through manipulation of resource availability and agent personalities, we observe how different agents engage in the competition and adapt their strategies. The use of LLM agents in game theory research offers significant advantages, including simulating realistic behavior, providing a controlled, scalable, and reproducible environment. Our work highlights the potential of LLM agents in enhancing the understanding of strategic decision-making within complex socioeconomic contexts. All codes are available at https://github.com/microsoft/Alympics

SCALE: Synergized Collaboration of Asymmetric Language Translation Engines

In this paper, we introduce SCALE, a collaborative framework that connects compact Specialized Translation Models (STMs) and general-purpose Large Language Models (LLMs) as one unified translation engine. By introducing translation from STM into the triplet in-context demonstrations, SCALE unlocks refinement and pivoting ability of LLM, thus mitigating language bias of LLM and parallel data bias of STM, enhancing LLM speciality without sacrificing generality, and facilitating continual learning without expensive LLM fine-tuning. Our comprehensive experiments show that SCALE significantly outperforms both few-shot LLMs (GPT-4) and specialized models (NLLB) in challenging low-resource settings. Moreover, in Xhosa to English translation, SCALE experiences consistent improvement by a 4 BLEURT score without tuning LLM and surpasses few-shot GPT-4 by 2.5 COMET score and 3.8 BLEURT score when equipped with a compact model consisting of merely 600M parameters. SCALE could also effectively exploit the existing language bias of LLMs by using an English-centric STM as a pivot for translation between any language pairs, outperforming few-shot GPT-4 by an average of 6 COMET points across eight translation directions. Furthermore we provide an in-depth analysis of SCALE's robustness, translation characteristics, and latency costs, providing solid foundation for future studies exploring the potential synergy between LLMs and more specialized, task-specific models.

Unleashing Cognitive Synergy in Large Language Models: A Task-Solving Agent through Multi-Persona Self-Collaboration

Human intelligence thrives on the concept of cognitive synergy, where collaboration and information integration among different cognitive processes yield superior outcomes compared to individual cognitive processes in isolation. Although Large Language Models (LLMs) have demonstrated promising performance as general task-solving agents, they still struggle with tasks that require intensive domain knowledge and complex reasoning. In this work, we propose Solo Performance Prompting (SPP), which transforms a single LLM into a cognitive synergist by engaging in multi-turn self-collaboration with multiple personas. A cognitive synergist refers to an intelligent agent that collaborates with multiple minds, combining their individual strengths and knowledge, to enhance problem-solving and overall performance in complex tasks. By dynamically identifying and simulating different personas based on task inputs, SPP unleashes the potential of cognitive synergy in LLMs. We have discovered that assigning multiple, fine-grained personas in LLMs elicits better problem-solving abilities compared to using a single or fixed number of personas. We evaluate SPP on three challenging tasks: Trivia Creative Writing, Codenames Collaborative, and Logic Grid Puzzle, encompassing both knowledge-intensive and reasoning-intensive types. Unlike previous works, such as Chain-of-Thought, that solely enhance the reasoning abilities in LLMs, SPP effectively elicits internal knowledge acquisition abilities, reduces hallucination, and maintains strong reasoning capabilities. Code, data, and prompts can be found at: https://github.com/MikeWangWZHL/Solo-Performance-Prompting.git.

In-context Autoencoder for Context Compression in a Large Language Model

We propose the In-context Autoencoder (ICAE) for context compression in a large language model (LLM). The ICAE has two modules: a learnable encoder adapted with LoRA from an LLM for compressing a long context into a limited number of memory slots, and a fixed decoder which is the target LLM that can condition on the memory slots for various purposes. We first pretrain the ICAE using both autoencoding and language modeling objectives on massive text data, enabling it to generate memory slots that accurately and comprehensively represent the original context. Then, we fine-tune the pretrained ICAE on a small amount of instruct data to enhance its interaction with various prompts for producing desirable responses. Our experimental results demonstrate that the ICAE learned with our proposed pretraining and fine-tuning paradigm can effectively produce memory slots with $4\times$ context compression, which can be well conditioned on by the target LLM to respond to various prompts. The promising results demonstrate significant implications of the ICAE for its novel approach to the long context problem and its potential to reduce computation and memory overheads for LLM inference in practice, suggesting further research effort in context management for an LLM. Our code and data will be released shortly.

Smart Word Suggestions for Writing Assistance

Enhancing word usage is a desired feature for writing assistance. To further advance research in this area, this paper introduces "Smart Word Suggestions" (SWS) task and benchmark. Unlike other works, SWS emphasizes end-to-end evaluation and presents a more realistic writing assistance scenario. This task involves identifying words or phrases that require improvement and providing substitution suggestions. The benchmark includes human-labeled data for testing, a large distantly supervised dataset for training, and the framework for evaluation. The test data includes 1,000 sentences written by English learners, accompanied by over 16,000 substitution suggestions annotated by 10 native speakers. The training dataset comprises over 3.7 million sentences and 12.7 million suggestions generated through rules. Our experiments with seven baselines demonstrate that SWS is a challenging task. Based on experimental analysis, we suggest potential directions for future research on SWS. The dataset and related codes is available at https://github.com/microsoft/SmartWordSuggestions.

Low-code LLM: Visual Programming over LLMs

Effectively utilizing LLMs for complex tasks is challenging, often involving a time-consuming and uncontrollable prompt engineering process. This paper introduces a novel human-LLM interaction framework, Low-code LLM. It incorporates six types of simple low-code visual programming interactions, all supported by clicking, dragging, or text editing, to achieve more controllable and stable responses. Through visual interaction with a graphical user interface, users can incorporate their ideas into the workflow without writing trivial prompts. The proposed Low-code LLM framework consists of a Planning LLM that designs a structured planning workflow for complex tasks, which can be correspondingly edited and confirmed by users through low-code visual programming operations, and an Executing LLM that generates responses following the user-confirmed workflow. We highlight three advantages of the low-code LLM: controllable generation results, user-friendly human-LLM interaction, and broadly applicable scenarios. We demonstrate its benefits using four typical applications. By introducing this approach, we aim to bridge the gap between humans and LLMs, enabling more effective and efficient utilization of LLMs for complex tasks. Our system will be soon publicly available at LowCodeLLM.

Inference with Reference: Lossless Acceleration of Large Language Models

We propose LLMA, an LLM accelerator to losslessly speed up Large Language Model (LLM) inference with references. LLMA is motivated by the observation that there are abundant identical text spans between the decoding result by an LLM and the reference that is available in many real world scenarios (e.g., retrieved documents). LLMA first selects a text span from the reference and copies its tokens to the decoder and then efficiently checks the tokens' appropriateness as the decoding result in parallel within one decoding step. The improved computational parallelism allows LLMA to achieve over 2x speed-up for LLMs with identical generation results as greedy decoding in many practical generation scenarios where significant overlap between in-context reference and outputs exists (e.g., search engines and multi-turn conversations).

Semiparametric Language Models Are Scalable Continual Learners

Semiparametric language models (LMs) have shown promise in continuously learning from new text data by combining a parameterized neural LM with a growable non-parametric memory for memorizing new content. However, conventional semiparametric LMs will finally become prohibitive for computing and storing if they are applied to continual learning over streaming data, because the non-parametric memory grows linearly with the amount of data they learn from over time. To address the issue of scalability, we present a simple and intuitive approach called Selective Memorization (SeMem), which only memorizes difficult samples that the model is likely to struggle with. We demonstrate that SeMem improves the scalability of semiparametric LMs for continual learning over streaming data in two ways: (1) data-wise scalability: as the model becomes stronger through continual learning, it will encounter fewer difficult cases that need to be memorized, causing the growth of the non-parametric memory to slow down over time rather than growing at a linear rate with the size of training data; (2) model-wise scalability: SeMem allows a larger model to memorize fewer samples than its smaller counterpart because it is rarer for a larger model to encounter incomprehensible cases, resulting in a non-parametric memory that does not scale linearly with model size. We conduct extensive experiments in language modeling and downstream tasks to test SeMem's results, showing SeMem enables a semiparametric LM to be a scalable continual learner with little forgetting.

MB-DECTNet: A Model-Based Unrolled Network for Accurate 3D DECT Reconstruction

Numerous dual-energy CT (DECT) techniques have been developed in the past few decades. Dual-energy CT (DECT) statistical iterative reconstruction (SIR) has demonstrated its potential for reducing noise and increasing accuracy. Our lab proposed a joint statistical DECT algorithm for stopping power estimation and showed that it outperforms competing image-based material-decomposition methods. However, due to its slow convergence and the high computational cost of projections, the elapsed time of 3D DECT SIR is often not clinically acceptable. Therefore, to improve its convergence, we have embedded DECT SIR into a deep learning model-based unrolled network for 3D DECT reconstruction (MB-DECTNet) that can be trained in an end-to-end fashion. This deep learning-based method is trained to learn the shortcuts between the initial conditions and the stationary points of iterative algorithms while preserving the unbiased estimation property of model-based algorithms. MB-DECTNet is formed by stacking multiple update blocks, each of which consists of a data consistency layer (DC) and a spatial mixer layer, where the spatial mixer layer is the shrunken U-Net, and the DC layer is a one-step update of an arbitrary traditional iterative method. Although the proposed network can be combined with numerous iterative DECT algorithms, we demonstrate its performance with the dual-energy alternating minimization (DEAM). The qualitative result shows that MB-DECTNet with DEAM significantly reduces noise while increasing the resolution of the test image. The quantitative result shows that MB-DECTNet has the potential to estimate attenuation coefficients accurately as traditional statistical algorithms but with a much lower computational cost.