Large language models (LLMs) have shown remarkable reasoning capabilities given chain-of-thought prompts (examples with intermediate reasoning steps). Existing benchmarks measure reasoning ability indirectly, by evaluating accuracy on downstream tasks such as mathematical reasoning. However, it is unclear how these models obtain the answers and whether they rely on simple heuristics rather than the generated chain-of-thought. To enable systematic exploration of the reasoning ability of LLMs, we present a new synthetic question-answering dataset called PrOntoQA, where each example is generated from a synthetic world model represented in first-order logic. This allows us to parse the generated chain-of-thought into symbolic proofs for formal analysis. Our analysis on InstructGPT and GPT-3 shows that LLMs are quite capable of making correct individual deduction steps, and so are generally capable of reasoning, even in fictional contexts. However, they have difficulty with proof planning: When multiple valid deduction steps are available, they are not able to systematically explore the different options.
相關內容
New emerging technologies powered by Artificial Intelligence (AI) have the potential to disruptively transform our societies for the better. In particular, data-driven learning approaches (i.e., Machine Learning (ML)) have been a true revolution in the advancement of multiple technologies in various application domains. But at the same time there is growing concern about certain intrinsic characteristics of these methodologies that carry potential risks to both safety and fundamental rights. Although there are mechanisms in the adoption process to minimize these risks (e.g., safety regulations), these do not exclude the possibility of harm occurring, and if this happens, victims should be able to seek compensation. Liability regimes will therefore play a key role in ensuring basic protection for victims using or interacting with these systems. However, the same characteristics that make AI systems inherently risky, such as lack of causality, opacity, unpredictability or their self and continuous learning capabilities, may lead to considerable difficulties when it comes to proving causation. This paper presents three case studies, as well as the methodology to reach them, that illustrate these difficulties. Specifically, we address the cases of cleaning robots, delivery drones and robots in education. The outcome of the proposed analysis suggests the need to revise liability regimes to alleviate the burden of proof on victims in cases involving AI technologies.
Transition systems are often used to describe the behaviour of software systems. If viewed as a graph then, at their most basic level, vertices correspond to the states of a program and each edge represents a transition between states via the (atomic) action labelled. In this setting, systems are thought to be consistent so that at each state formulas are evaluated as either True or False. On the other hand, when a structure of this sort - for example a map where states represent locations, some local properties are known and labelled transitions represent information available about different routes - is built resorting to multiple sources of information, it is common to find inconsistent or incomplete information regarding what holds at each state, both at the level of propositional variables and transitions. This paper aims at bringing together Belnap's four values, Dynamic Logic and hybrid machinery such as nominals and the satisfaction operator, so that reasoning is still possible in face of contradicting evidence. Proof-theory for this new logic is explored by means of a terminating, sound and complete tableaux system.
While the majority of autonomous driving research has concentrated on everyday driving scenarios, further safety and performance improvements of autonomous vehicles require a focus on extreme driving conditions. In this context, autonomous racing is a new area of research that has been attracting considerable interest recently. Due to the fact that a vehicle is driven by its perception, planning, and control limits during racing, numerous research and development issues arise. This paper provides a comprehensive overview of the autonomous racing system built by team KAIST for the Indy Autonomous Challenge (IAC). Our autonomy stack consists primarily of a multi-modal perception module, a high-speed overtaking planner, a resilient control stack, and a system status manager. We present the details of all components of our autonomy solution, including algorithms, implementation, and unit test results. In addition, this paper outlines the design principles and the results of a systematical analysis. Even though our design principles are derived from the unique application domain of autonomous racing, they can also be applied to a variety of safety-critical, high-cost-of-failure robotics applications. The proposed system was integrated into a full-scale autonomous race car (Dallara AV-21) and field-tested extensively. As a result, team KAIST was one of three teams who qualified and participated in the official IAC race events without any accidents. Our proposed autonomous system successfully completed all missions, including overtaking at speeds of around $220 km/h$ in the IAC@CES2022, the world's first autonomous 1:1 head-to-head race.
Automaton-based representations of task knowledge play an important role in control and planning for sequential decision-making problems. However, obtaining the high-level task knowledge required to build such automata is often difficult. Meanwhile, large-scale generative language models (GLMs) can automatically generate relevant task knowledge. However, the textual outputs from GLMs cannot be formally verified or used for sequential decision-making. We propose a novel algorithm named GLM2FSA, which constructs a finite state automaton (FSA) encoding high-level task knowledge from a brief natural-language description of the task goal. GLM2FSA first sends queries to a GLM to extract task knowledge in textual form, and then it builds an FSA to represent this text-based knowledge. The proposed algorithm thus fills the gap between natural-language task descriptions and automaton-based representations, and the constructed FSA can be formally verified against user-defined specifications. We accordingly propose a method to iteratively refine the queries to the GLM based on the outcomes, e.g., counter-examples, from verification. We demonstrate GLM2FSA's ability to build and refine automaton-based representations of everyday tasks (e.g., crossing a road or making a phone call), and also of tasks that require highly-specialized knowledge (e.g., executing secure multi-party computation).
Formally verifying software properties is a highly desirable but labor-intensive task. Recent work has developed methods to automate formal verification using proof assistants, such as Coq and Isabelle/HOL, e.g., by training a model to predict one proof step at a time, and using that model to search through the space of possible proofs. This paper introduces a new method to automate formal verification: We use large language models, trained on natural language text and code and fine-tuned on proofs, to generate whole proofs for theorems at once, rather than one step at a time. We combine this proof generation model with a fine-tuned repair model to repair generated proofs, further increasing proving power. As its main contributions, this paper demonstrates for the first time that: (1) Whole-proof generation using transformers is possible and is as effective as search-based techniques without requiring costly search. (2) Giving the learned model additional context, such as a prior failed proof attempt and the ensuing error message, results in proof repair and further improves automated proof generation. (3) We establish a new state of the art for fully automated proof synthesis. We reify our method in a prototype, Baldur, and evaluate it on a benchmark of 6,336 Isabelle/HOL theorems and their proofs. In addition to empirically showing the effectiveness of whole-proof generation, repair, and added context, we show that Baldur improves on the state-of-the-art tool, Thor, by automatically generating proofs for an additional 8.7% of the theorems. Together, Baldur and Thor can prove 65.7% of the theorems fully automatically. This paper paves the way for new research into using large language models for automating formal verification.
Large language models (LLMs) can perform complex reasoning in few- and zero-shot settings by generating intermediate chain of thought (CoT) reasoning steps. Further, each reasoning step can rely on external tools to support computation beyond the core LLM capabilities (e.g. search/running code). Prior work on CoT prompting and tool use typically requires hand-crafting task-specific demonstrations and carefully scripted interleaving of model generations with tool use. We introduce Automatic Reasoning and Tool-use (ART), a framework that uses frozen LLMs to automatically generate intermediate reasoning steps as a program. Given a new task to solve, ART selects demonstrations of multi-step reasoning and tool use from a task library. At test time, ART seamlessly pauses generation whenever external tools are called, and integrates their output before resuming generation. ART achieves a substantial improvement over few-shot prompting and automatic CoT on unseen tasks in the BigBench and MMLU benchmarks, and matches performance of hand-crafted CoT prompts on a majority of these tasks. ART is also extensible, and makes it easy for humans to improve performance by correcting errors in task-specific programs or incorporating new tools, which we demonstrate by drastically improving performance on select tasks with minimal human intervention.
Reasoning is a fundamental aspect of human intelligence that plays a crucial role in activities such as problem solving, decision making, and critical thinking. In recent years, large language models (LLMs) have made significant progress in natural language processing, and there is observation that these models may exhibit reasoning abilities when they are sufficiently large. However, it is not yet clear to what extent LLMs are capable of reasoning. This paper provides a comprehensive overview of the current state of knowledge on reasoning in LLMs, including techniques for improving and eliciting reasoning in these models, methods and benchmarks for evaluating reasoning abilities, findings and implications of previous research in this field, and suggestions on future directions. Our aim is to provide a detailed and up-to-date review of this topic and stimulate meaningful discussion and future work.
Knowledge enhanced pre-trained language models (K-PLMs) are shown to be effective for many public tasks in the literature but few of them have been successfully applied in practice. To address this problem, we propose K-AID, a systematic approach that includes a low-cost knowledge acquisition process for acquiring domain knowledge, an effective knowledge infusion module for improving model performance, and a knowledge distillation component for reducing the model size and deploying K-PLMs on resource-restricted devices (e.g., CPU) for real-world application. Importantly, instead of capturing entity knowledge like the majority of existing K-PLMs, our approach captures relational knowledge, which contributes to better-improving sentence-level text classification and text matching tasks that play a key role in question answering (QA). We conducted a set of experiments on five text classification tasks and three text matching tasks from three domains, namely E-commerce, Government, and Film&TV, and performed online A/B tests in E-commerce. Experimental results show that our approach is able to achieve substantial improvement on sentence-level question answering tasks and bring beneficial business value in industrial settings.
AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks. We call these models foundation models to underscore their critically central yet incomplete character. This report provides a thorough account of the opportunities and risks of foundation models, ranging from their capabilities (e.g., language, vision, robotics, reasoning, human interaction) and technical principles(e.g., model architectures, training procedures, data, systems, security, evaluation, theory) to their applications (e.g., law, healthcare, education) and societal impact (e.g., inequity, misuse, economic and environmental impact, legal and ethical considerations). Though foundation models are based on standard deep learning and transfer learning, their scale results in new emergent capabilities,and their effectiveness across so many tasks incentivizes homogenization. Homogenization provides powerful leverage but demands caution, as the defects of the foundation model are inherited by all the adapted models downstream. Despite the impending widespread deployment of foundation models, we currently lack a clear understanding of how they work, when they fail, and what they are even capable of due to their emergent properties. To tackle these questions, we believe much of the critical research on foundation models will require deep interdisciplinary collaboration commensurate with their fundamentally sociotechnical nature.
This paper surveys and organizes research works in a new paradigm in natural language processing, which we dub "prompt-based learning". Unlike traditional supervised learning, which trains a model to take in an input x and predict an output y as P(y|x), prompt-based learning is based on language models that model the probability of text directly. To use these models to perform prediction tasks, the original input x is modified using a template into a textual string prompt x' that has some unfilled slots, and then the language model is used to probabilistically fill the unfilled information to obtain a final string x, from which the final output y can be derived. This framework is powerful and attractive for a number of reasons: it allows the language model to be pre-trained on massive amounts of raw text, and by defining a new prompting function the model is able to perform few-shot or even zero-shot learning, adapting to new scenarios with few or no labeled data. In this paper we introduce the basics of this promising paradigm, describe a unified set of mathematical notations that can cover a wide variety of existing work, and organize existing work along several dimensions, e.g.the choice of pre-trained models, prompts, and tuning strategies. To make the field more accessible to interested beginners, we not only make a systematic review of existing works and a highly structured typology of prompt-based concepts, but also release other resources, e.g., a website //pretrain.nlpedia.ai/ including constantly-updated survey, and paperlist.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191