Large Language Models (LLMs) have shown promise as intelligent agents in interactive decision-making tasks. Traditional approaches often depend on meticulously designed prompts, high-quality examples, or additional reward models for in-context learning, supervised fine-tuning, or RLHF. Reinforcement learning (RL) presents a dynamic alternative for LLMs to overcome these dependencies by engaging directly with task-specific environments. Nonetheless, it faces significant hurdles: 1) instability stemming from the exponentially vast action space requiring exploration; 2) challenges in assigning token-level credit based on action-level reward signals, resulting in discord between maximizing rewards and accurately modeling corpus data. In response to these challenges, we introduce Entropy-Regularized Token-level Policy Optimization (ETPO), an entropy-augmented RL method tailored for optimizing LLMs at the token level. At the heart of ETPO is our novel per-token soft Bellman update, designed to harmonize the RL process with the principles of language modeling. This methodology decomposes the Q-function update from a coarse action-level view to a more granular token-level perspective, backed by theoretical proof of optimization consistency. Crucially, this decomposition renders linear time complexity in action exploration. We assess the effectiveness of ETPO within a simulated environment that models data science code generation as a series of multi-step interactive tasks; results underline ETPO's potential as a robust method for refining the interactive decision-making capabilities of language agents. For a more detailed preliminary work describing our motivation for token-level decomposition and applying it in PPO methods, please refer to arXiv:2405.15821.
Several attempts have been made to implement text command control for game agents. However, current technologies are limited to processing predefined format commands. This paper proposes a pioneering text command control system for a game agent that can understand natural language commands expressed in free-form. The proposed system uses a large language model (LLM) for code generation to interpret and transform natural language commands into behavior branch, a proposed knowledge expression based on behavior trees, which facilitates execution by the game agent. This study conducted empirical validation within a game environment that simulates a Pok\'emon game and involved multiple participants. The results confirmed the system's ability to understand and carry out natural language commands, representing a noteworthy in the realm of real-time language interactive game agents. Notice for the use of this material. The copyright of this material is retained by the Japanese Society for Artificial Intelligence (JSAI). This material is published here with the agreement of JSAI. Please be complied with Copyright Law of Japan if any users wish to reproduce, make derivative work, distribute or make available to the public any part or whole thereof. All Rights Reserved, Copyright (C) The Japanese Society for Artificial Intelligence.
Common self-improvement approaches for large language models (LLMs), such as STaR, iteratively fine-tune LLMs on self-generated solutions to improve their problem-solving ability. However, these approaches discard the large amounts of incorrect solutions generated during this process, potentially neglecting valuable information in such solutions. To address this shortcoming, we propose V-STaR that utilizes both the correct and incorrect solutions generated during the self-improvement process to train a verifier using DPO that judges correctness of model-generated solutions. This verifier is used at inference time to select one solution among many candidate solutions. Running V-STaR for multiple iterations results in progressively better reasoners and verifiers, delivering a 4% to 17% test accuracy improvement over existing self-improvement and verification approaches on common code generation and math reasoning benchmarks with LLaMA2 models.
Large language models and multimodal large language models have revolutionized artificial intelligence recently. An increasing number of regions are now embracing these advanced technologies. Within this context, robot coding education is garnering increasing attention. To teach young children how to code and compete in robot challenges, large language models are being utilized for robot code explanation, generation, and modification. In this paper, we highlight an important trend in robot coding education. We test several mainstream large language models on both traditional coding tasks and the more challenging task of robot code generation, which includes block diagrams. Our results show that GPT-4V outperforms other models in all of our tests but struggles with generating block diagram images.
Large Language Models' success on text generation has also made them better at code generation and coding tasks. While a lot of work has demonstrated their remarkable performance on tasks such as code completion and editing, it is still unclear as to why. We help bridge this gap by exploring to what degree auto-regressive models understand the logical constructs of the underlying programs. We propose Counterfactual Analysis for Programming Concept Predicates (CACP) as a counterfactual testing framework to evaluate whether Large Code Models understand programming concepts. With only black-box access to the model, we use CACP to evaluate ten popular Large Code Models for four different programming concepts. Our findings suggest that current models lack understanding of concepts such as data flow and control flow.
Large pre-trained language models have recently been expanded and applied to programming language tasks with great success, often through further pre-training of a strictly-natural language model--where training sequences typically contain both natural and (linearised) programming language. Such approaches effectively map both modalities of the sequence into the same embedding space. However, programming language keywords (e.g. "while") often have very strictly defined semantics. As such, transfer learning from their natural language usage may not necessarily be beneficial to their code application and vise versa. Assuming an already pre-trained language model, in this work we investigate how sequence tokens can be adapted and represented differently, depending on which modality they belong to, and to the ultimate benefit of the downstream task. We experiment with separating embedding spaces between modalities during further model pre-training with modality-relative training objectives. We focus on text-to-code generation and observe consistent improvements across two backbone models and two test sets, measuring pass@$k$ and a novel incremental variation.
We analyze the capabilities of Transformer language models in learning compositional discrete tasks. To this end, we evaluate training LLaMA models and prompting GPT-4 and Gemini on four tasks demanding to learn a composition of several discrete sub-tasks. In particular, we measure how well these models can reuse primitives observable in the sub-tasks to learn the composition task. Our results indicate that compositional learning in state-of-the-art Transformer language models is highly sample inefficient: LLaMA requires more data samples than relearning all sub-tasks from scratch to learn the compositional task; in-context prompting with few samples is unreliable and fails at executing the sub-tasks or correcting the errors in multi-round code generation. Further, by leveraging complexity theory, we support these findings with a theoretical analysis focused on the sample inefficiency of gradient descent in memorizing feedforward models. We open source our code at https://github.com/IBM/limitations-lm-algorithmic-compositional-learning.
Pre-trained Vision-Language Models (VLMs) are able to understand visual concepts, describe and decompose complex tasks into sub-tasks, and provide feedback on task completion. In this paper, we aim to leverage these capabilities to support the training of reinforcement learning (RL) agents. In principle, VLMs are well suited for this purpose, as they can naturally analyze image-based observations and provide feedback (reward) on learning progress. However, inference in VLMs is computationally expensive, so querying them frequently to compute rewards would significantly slowdown the training of an RL agent. To address this challenge, we propose a framework named Code as Reward (VLM-CaR). VLM-CaR produces dense reward functions from VLMs through code generation, thereby significantly reducing the computational burden of querying the VLM directly. We show that the dense rewards generated through our approach are very accurate across a diverse set of discrete and continuous environments, and can be more effective in training RL policies than the original sparse environment rewards.
Stratospheric balloons have emerged as an affordable and flexible alternative to traditional spacecrafts as they are implemented using commercial off-the-shelf (COTS) equipment without following strict methodologies. HERCCULES is a stratospheric balloon mission that aims to characterize the convective heat and radiative environment in the stratosphere. The purpose of this article is to present the HERCCULES onboard software (OBSW) whose design and complexity is comparable to that of satellite systems, since it must control about sixty COTS equipment using a single Raspberry Pi 4B as onboard computer and ensure the real-time requirements. Compared to similar systems, novel contributions are presented as the OBSW is developed following modelbased and component-based approaches using the TASTE toolchain from the European Space Agency (ESA) for automatic code generation. Besides, the OBSW is verified and validated following the ESA standards and the results obtained demonstrate the suitability and efficiency of the solution and the selected methodologies.
Despite the remarkable code generation abilities of large language models LLMs, they still face challenges in complex task handling. Robot development, a highly intricate field, inherently demands human involvement in task allocation and collaborative teamwork . To enhance robot development, we propose an innovative automated collaboration framework inspired by real-world robot developers. This framework employs multiple LLMs in distinct roles analysts, programmers, and testers. Analysts delve deep into user requirements, enabling programmers to produce precise code, while testers fine-tune the parameters based on user feedback for practical robot application. Each LLM tackles diverse, critical tasks within the development process. Clear collaboration rules emulate real world teamwork among LLMs. Analysts, programmers, and testers form a cohesive team overseeing strategy, code, and parameter adjustments . Through this framework, we achieve complex robot development without requiring specialized knowledge, relying solely on non experts participation.
This study evaluates the effectiveness of various large language models (LLMs) in performing tasks common among undergraduate computer science students. Although a number of research studies in the computing education community have explored the possibility of using LLMs for a variety of tasks, there is a lack of comprehensive research comparing different LLMs and evaluating which LLMs are most effective for different tasks. Our research systematically assesses some of the publicly available LLMs such as Google Bard, ChatGPT(3.5), GitHub Copilot Chat, and Microsoft Copilot across diverse tasks commonly encountered by undergraduate computer science students in India. These tasks include code explanation and documentation, solving class assignments, technical interview preparation, learning new concepts and frameworks, and email writing. Evaluation for these tasks was carried out by pre-final year and final year undergraduate computer science students and provides insights into the models' strengths and limitations. This study aims to guide students as well as instructors in selecting suitable LLMs for any specific task and offers valuable insights on how LLMs can be used constructively by students and instructors.
Code generation models have increasingly become integral to aiding software development. Although current research has thoroughly examined the correctness of the code produced by code generation models, a vital aspect that plays a pivotal role in green computing and sustainability efforts has often been neglected. This paper presents EffiBench, a benchmark with 1,000 efficiency-critical coding problems to assess the efficiency of code generated by code generation models. EffiBench contains a diverse set of LeetCode coding problems. Each problem is paired with an executable human-written canonical solution, which obtains the SOTA efficiency on the LeetCode solution leaderboard. With EffiBench, we empirically examine the ability of 42 large language models (35 open-source and 7 closed-source) to generate efficient code. Our evaluation results demonstrate that the efficiency of the code generated by LLMs is generally worse than the efficiency of human-written canonical solutions. For example, GPT-4 generated code has an average \textbf{3.12} times execution time that of the human-written canonical solutions. In the most extreme cases, the execution time and total memory usage of GPT-4 generated code are \textbf{13.89} and \textbf{43.92} times that of the canonical solutions. The source code of EffiBench is released on https://github.com/huangd1999/EffiBench. We also provide the LeaderBoard at https://huggingface.co/spaces/EffiBench/effibench-leaderboard.
Recent studies have shown that code language models at scale demonstrate significant performance gains on downstream tasks, i.e., code generation. However, most of the existing works on code representation learning train models at a hundred million parameter scale using very limited pretraining corpora. In this work, we fuel code representation learning with a vast amount of code data via a two-stage pretraining scheme. We first train the encoders via a mix that leverages both randomness in masking language modeling and the structure aspect of programming language. We then enhance the representations via contrastive learning with hard negative and hard positive constructed in an unsupervised manner. We establish an off-the-shelf encoder model that persistently outperforms the existing models on a wide variety of downstream tasks by large margins. To comprehend the factors contributing to successful code representation learning, we conduct detailed ablations and share our findings on (i) a customized and effective token-level denoising scheme for source code; (ii) the importance of hard negatives and hard positives; (iii) how the proposed bimodal contrastive learning boost the cross-lingual semantic search performance; and (iv) how the pretraining schemes decide the downstream task performance scales with the model size.
Existing large language models (LLMs) for register transfer level code generation face challenges like compilation failures and suboptimal power, performance, and area (PPA) efficiency. This is due to the lack of PPA awareness in conventional transformer decoding algorithms. In response, we present an automated transformer decoding algorithm that integrates Monte Carlo tree-search for lookahead, guiding the transformer to produce compilable, functionally correct, and PPA-optimized code. Empirical evaluation with a fine-tuned language model on RTL codesets shows that our proposed technique consistently generates functionally correct code compared to prompting-only methods and effectively addresses the PPA-unawareness drawback of naive large language models. For the largest design generated by the state-of-the-art LLM (16-bit adder), our technique can achieve a 31.8% improvement in the area-delay product.
With the rapid advance of machine learning (ML) technology, large language models (LLMs) are increasingly explored as an intelligent tool to generate program code from natural language specifications. However, existing evaluations of LLMs have focused on their capabilities in comparison with humans. It is desirable to evaluate their usability when deciding on whether to use a LLM in software production. This paper proposes a user centric method for this purpose. It includes metadata in the test cases of a benchmark to describe their usages, conducts testing in a multi-attempt process that mimics the uses of LLMs, measures LLM generated solutions on a set of quality attributes that reflect usability, and evaluates the performance based on user experiences in the uses of LLMs as a tool. The paper also reports a case study with the method in the evaluation of ChatGPT's usability as a code generation tool for the R programming language. Our experiments demonstrated that ChatGPT is highly useful for generating R program code although it may fail on hard programming tasks. The user experiences are good with overall average number of attempts being 1.61 and the average time of completion being 47.02 seconds. Our experiments also found that the weakest aspect of usability is conciseness, which has a score of 3.80 out of 5.
The advancement of large language models (LLMs) has significantly propelled the field of code generation. Previous work integrated reinforcement learning (RL) with compiler feedback for exploring the output space of LLMs to enhance code generation quality. However, the lengthy code generated by LLMs in response to complex human requirements makes RL exploration a challenge. Also, since the unit tests may not cover the complicated code, optimizing LLMs by using these unexecuted code snippets is ineffective. To tackle these challenges, we introduce StepCoder, a novel RL framework for code generation, consisting of two main components: CCCS addresses the exploration challenge by breaking the long sequences code generation task into a Curriculum of Code Completion Subtasks, while FGO only optimizes the model by masking the unexecuted code segments to provide Fine-Grained Optimization. In addition, we furthermore construct the APPS+ dataset for RL training, which is manually verified to ensure the correctness of unit tests. Experimental results show that our method improves the ability to explore the output space and outperforms state-of-the-art approaches in corresponding benchmarks. Our dataset APPS+ and StepCoder are available online.
Tokenization is an understudied and often neglected component of modern LLMs. Most published works use a single tokenizer for all experiments, often borrowed from another model, without performing ablations or analysis to optimize tokenization. Moreover, the tokenizer is generally kept unchanged when fine-tuning a base model. In this paper, we show that the size, pre-tokenization regular expression, and training data of a tokenizer can significantly impact the model's generation speed, effective context size, memory usage, and downstream performance. We train specialized Byte-Pair Encoding code tokenizers, and conduct extensive ablations on the impact of tokenizer design on the performance of LLMs for code generation tasks such as HumanEval and MBPP, and provide recommendations for tokenizer hyper-parameters selection and switching the tokenizer in a pre-trained LLM. We perform our experiments on models trained from scratch and from pre-trained models, verifying their applicability to a wide range of use-cases. We find that when fine-tuning on more than 50 billion tokens, we can specialize the tokenizer of a pre-trained LLM to obtain large gains in generation speed and effective context size.
The advent of foundation models has revolutionized the fields of natural language processing and computer vision, paving the way for their application in autonomous driving (AD). This survey presents a comprehensive review of more than 40 research papers, demonstrating the role of foundation models in enhancing AD. Large language models contribute to planning and simulation in AD, particularly through their proficiency in reasoning, code generation and translation. In parallel, vision foundation models are increasingly adapted for critical tasks such as 3D object detection and tracking, as well as creating realistic driving scenarios for simulation and testing. Multi-modal foundation models, integrating diverse inputs, exhibit exceptional visual understanding and spatial reasoning, crucial for end-to-end AD. This survey not only provides a structured taxonomy, categorizing foundation models based on their modalities and functionalities within the AD domain but also delves into the methods employed in current research. It identifies the gaps between existing foundation models and cutting-edge AD approaches, thereby charting future research directions and proposing a roadmap for bridging these gaps.
$ $Large Language Models (LLMs) are being increasingly utilized in various applications, with code generations being a notable example. While previous research has shown that LLMs have the capability to generate both secure and insecure code, the literature does not take into account what factors help generate secure and effective code. Therefore in this paper we focus on identifying and understanding the conditions and contexts in which LLMs can be effectively and safely deployed in real-world scenarios to generate quality code. We conducted a comparative analysis of four advanced LLMs--GPT-3.5 and GPT-4 using ChatGPT and Bard and Gemini from Google--using 9 separate tasks to assess each model's code generation capabilities. We contextualized our study to represent the typical use cases of a real-life developer employing LLMs for everyday tasks as work. Additionally, we place an emphasis on security awareness which is represented through the use of two distinct versions of our developer persona. In total, we collected 61 code outputs and analyzed them across several aspects: functionality, security, performance, complexity, and reliability. These insights are crucial for understanding the models' capabilities and limitations, guiding future development and practical applications in the field of automated code generation.
Large language models (LLMs)such as ChatGPT have significantly advanced the field of Natural Language Processing (NLP). This trend led to the development of code-based large language models such as StarCoder, WizardCoder, and CodeLlama, which are trained extensively on vast repositories of code and programming languages. While the generic abilities of these code LLMs are useful for many programmers in tasks like code generation, the area of high-performance computing (HPC) has a narrower set of requirements that make a smaller and more domain-specific model a smarter choice. This paper presents OMPGPT, a novel domain-specific model meticulously designed to harness the inherent strengths of language models for OpenMP pragma generation. Furthermore, we leverage prompt engineering techniques from the NLP domain to create Chain-of-OMP, an innovative strategy designed to enhance OMPGPT's effectiveness. Our extensive evaluations demonstrate that OMPGPT outperforms existing large language models specialized in OpenMP tasks and maintains a notably smaller size, aligning it more closely with the typical hardware constraints of HPC environments. We consider our contribution as a pivotal bridge, connecting the advantage of language models with the specific demands of HPC tasks.