The performance on Large Language Models (LLMs) on existing reasoning benchmarks has shot up considerably over the past years. In response, we present JEEBench, a considerably more challenging benchmark dataset for evaluating the problem solving abilities of LLMs. We curate 450 challenging pre-engineering mathematics, physics and chemistry problems from the IIT JEE-Advanced exam. Long-horizon reasoning on top of deep in-domain knowledge is essential for solving problems in this benchmark. Our evaluation on the GPT series of models reveals that although performance improves with newer models, the best being GPT-4, the highest performance, even after using techniques like Self-Consistency and Chain-of-Thought prompting is less than 40 percent. Our analysis demonstrates that errors in algebraic manipulation and failure in retrieving relevant domain specific concepts are primary contributors to GPT4's low performance. Given the challenging nature of the benchmark, we hope that it can guide future research in problem solving using LLMs. Our code and dataset is available here.
相關內容
Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, including position, verbosity, and self-enhancement biases, as well as limited reasoning ability, and propose solutions to mitigate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-bench, a multi-turn question set; and Chatbot Arena, a crowdsourced battle platform. Our results reveal that strong LLM judges like GPT-4 can match both controlled and crowdsourced human preferences well, achieving over 80\% agreement, the same level of agreement between humans. Hence, LLM-as-a-judge is a scalable and explainable way to approximate human preferences, which are otherwise very expensive to obtain. Additionally, we show our benchmark and traditional benchmarks complement each other by evaluating several variants of LLaMA and Vicuna. We will publicly release MT-bench questions, 3K expert votes, and 30K conversations with human preferences from Chatbot Arena.
Methods to generate text from structured data have advanced significantly in recent years, primarily due to fine-tuning of pre-trained language models on large datasets. However, such models can fail to produce output faithful to the input data, particularly on out-of-domain data. Sufficient annotated data is often not available for specific domains, leading us to seek an unsupervised approach to improve the faithfulness of output text. Since the problem is fundamentally one of consistency between the representations of the structured data and text, we evaluate the effectiveness of cycle training in this work. Cycle training uses two models which are inverses of each other: one that generates text from structured data, and one which generates the structured data from natural language text. We show that cycle training, when initialized with a small amount of supervised data (100 samples in our case), achieves nearly the same performance as fully supervised approaches for the data-to-text generation task on the WebNLG, E2E, WTQ, and WSQL datasets. We perform extensive empirical analysis with automated evaluation metrics and a newly designed human evaluation schema to reveal different cycle training strategies' effectiveness of reducing various types of generation errors. Our code is publicly available at //github.com/Edillower/CycleNLG.
Benchmarks are among the main drivers of progress in software engineering research, especially in software testing and debugging. However, current benchmarks in this field could be better suited for specific research tasks, as they rely on weak system oracles like crash detection, come with few unit tests only, need more elaborative research, or cannot verify the outcome of system tests. Our Tests4Py benchmark addresses these issues. It is derived from the popular BugsInPy benchmark, including 30 bugs from 5 real-world Python applications. Each subject in Tests4Py comes with an oracle to verify the functional correctness of system inputs. Besides, it enables the generation of system tests and unit tests, allowing for qualitative studies by investigating essential aspects of test sets and extensive evaluations. These opportunities make Tests4Py a next-generation benchmark for research in test generation, debugging, and automatic program repair.
We observe that pre-trained large language models (LLMs) are capable of autoregressively completing complex token sequences -- from arbitrary ones procedurally generated by probabilistic context-free grammars (PCFG), to more rich spatial patterns found in the Abstract Reasoning Corpus (ARC), a general AI benchmark, prompted in the style of ASCII art. Surprisingly, pattern completion proficiency can be partially retained even when the sequences are expressed using tokens randomly sampled from the vocabulary. These results suggest that without any additional training, LLMs can serve as general sequence modelers, driven by in-context learning. In this work, we investigate how these zero-shot capabilities may be applied to problems in robotics -- from extrapolating sequences of numbers that represent states over time to complete simple motions, to least-to-most prompting of reward-conditioned trajectories that can discover and represent closed-loop policies (e.g., a stabilizing controller for CartPole). While difficult to deploy today for real systems due to latency, context size limitations, and compute costs, the approach of using LLMs to drive low-level control may provide an exciting glimpse into how the patterns among words could be transferred to actions.
The transformative influence of Large Language Models (LLMs) is profoundly reshaping the Artificial Intelligence (AI) technology domain. Notably, ChatGPT distinguishes itself within these models, demonstrating remarkable performance in multi-turn conversations and exhibiting code proficiency across an array of languages. In this paper, we carry out a comprehensive evaluation of ChatGPT's coding capabilities based on what is to date the largest catalog of coding challenges. Our focus is on the python programming language and problems centered on data structures and algorithms, two topics at the very foundations of Computer Science. We evaluate ChatGPT for its ability to generate correct solutions to the problems fed to it, its code quality, and nature of run-time errors thrown by its code. Where ChatGPT code successfully executes, but fails to solve the problem at hand, we look into patterns in the test cases passed in order to gain some insights into how wrong ChatGPT code is in these kinds of situations. To infer whether ChatGPT might have directly memorized some of the data that was used to train it, we methodically design an experiment to investigate this phenomena. Making comparisons with human performance whenever feasible, we investigate all the above questions from the context of both its underlying learning models (GPT-3.5 and GPT-4), on a vast array sub-topics within the main topics, and on problems having varying degrees of difficulty.
Large language models (LLMs) have significantly advanced the field of natural language processing, with GPT models at the forefront. While their remarkable performance spans a range of tasks, adapting LLMs for real-world business scenarios still poses challenges warranting further investigation. This paper presents an empirical analysis aimed at bridging the gap in adapting LLMs to practical use cases. To do that, we select the question answering (QA) task of insurance as a case study due to its challenge of reasoning. Based on the task we design a new model relied on LLMs which are empowered by additional knowledge extracted from insurance policy rulebooks and DBpedia. The additional knowledge helps LLMs to understand new concepts of insurance for domain adaptation. Preliminary results on two QA datasets show that knowledge enhancement significantly improves the reasoning ability of GPT-3.5 (55.80% and 57.83% in terms of accuracy). The analysis also indicates that existing public knowledge bases, e.g., DBPedia is beneficial for knowledge enhancement. Our findings reveal that the inherent complexity of business scenarios often necessitates the incorporation of domain-specific knowledge and external resources for effective problem-solving.
Supply chain operations traditionally involve a variety of complex decision making problems. Over the last few decades, supply chains greatly benefited from advances in computation, which allowed the transition from manual processing to automation and cost-effective optimization. Nonetheless, business operators still need to spend substantial efforts in \emph{explaining} and interpreting the optimization outcomes to stakeholders. Motivated by the recent advances in Large Language Models (LLMs), we study how this disruptive technology can help bridge the gap between supply chain automation and human comprehension and trust thereof. We design \name{} -- a framework that accepts as input queries in plain text, and outputs insights about the underlying optimization outcomes. Our framework does not forgo the state-of-the-art combinatorial optimization technology, but rather leverages it to quantitatively answer what-if scenarios (e.g., how would the cost change if we used supplier B instead of supplier A for a given demand?). Importantly, our design does not require sending proprietary data over to LLMs, which can be a privacy concern in some circumstances. We demonstrate the effectiveness of our framework on a real server placement scenario within Microsoft's cloud supply chain. Along the way, we develop a general evaluation benchmark, which can be used to evaluate the accuracy of the LLM output in other scenarios.
Large language models (LLMs) have shown remarkable abilities to generate code, however their ability to develop software for embedded systems, which requires cross-domain knowledge of hardware and software has not been studied. In this paper we systematically evaluate leading LLMs (GPT-3.5, GPT-4, PaLM 2) to assess their performance for embedded system development, study how human programmers interact with these tools, and develop an AI-based software engineering workflow for building embedded systems. We develop an an end-to-end hardware-in-the-loop evaluation platform for verifying LLM generated programs using sensor actuator pairs. We compare all three models with N=450 experiments and find surprisingly that GPT-4 especially shows an exceptional level of cross-domain understanding and reasoning, in some cases generating fully correct programs from a single prompt. In N=50 trials, GPT-4 produces functional I2C interfaces 66% of the time. GPT-4 also produces register-level drivers, code for LoRa communication, and context-specific power optimizations for an nRF52 program resulting in over 740x current reduction to 12.2 uA. We also characterize the models' limitations to develop a generalizable workflow for using LLMs in embedded system development. We evaluate the workflow with 15 users including novice and expert programmers. We find that our workflow improves productivity for all users and increases the success rate for building a LoRa environmental sensor from 25% to 100%, including for users with zero hardware or C/C++ experience.
The unprecedented performance of large language models (LLMs) necessitates improvements in evaluations. Rather than merely exploring the breadth of LLM abilities, we believe meticulous and thoughtful designs are essential to thorough, unbiased, and applicable evaluations. Given the importance of world knowledge to LLMs, we construct a Knowledge-oriented LLM Assessment benchmark (KoLA), in which we carefully design three crucial factors: (1) For ability modeling, we mimic human cognition to form a four-level taxonomy of knowledge-related abilities, covering $19$ tasks. (2) For data, to ensure fair comparisons, we use both Wikipedia, a corpus prevalently pre-trained by LLMs, along with continuously collected emerging corpora, aiming to evaluate the capacity to handle unseen data and evolving knowledge. (3) For evaluation criteria, we adopt a contrastive system, including overall standard scores for better numerical comparability across tasks and models and a unique self-contrast metric for automatically evaluating knowledge hallucination. We evaluate $21$ open-source and commercial LLMs and obtain some intriguing findings. The KoLA dataset and open-participation leaderboard are publicly released at //kola.xlore.cn and will be continuously updated to provide references for developing LLMs and knowledge-related systems.
Commonsense knowledge and commonsense reasoning are some of the main bottlenecks in machine intelligence. In the NLP community, many benchmark datasets and tasks have been created to address commonsense reasoning for language understanding. These tasks are designed to assess machines' ability to acquire and learn commonsense knowledge in order to reason and understand natural language text. As these tasks become instrumental and a driving force for commonsense research, this paper aims to provide an overview of existing tasks and benchmarks, knowledge resources, and learning and inference approaches toward commonsense reasoning for natural language understanding. Through this, our goal is to support a better understanding of the state of the art, its limitations, and future challenges.
小貼士
登錄享
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191