Recent large language models (LLMs) advancements sparked a growing research interest in tool assisted LLMs solving real-world challenges, which calls for comprehensive evaluation of tool-use capabilities. While previous works focused on either evaluating over stateless web services (RESTful API), based on a single turn user prompt, or an off-policy dialog trajectory, ToolSandbox includes stateful tool execution, implicit state dependencies between tools, a built-in user simulator supporting on-policy conversational evaluation and a dynamic evaluation strategy for intermediate and final milestones over arbitrary trajectory. We show that open source and proprietary models has a significant performance gap, and complex tasks like State Dependency, Canonicalization and Insufficient Information defined in ToolSandbox are challenging even the most capable SOTA LLMs, providing brand-new insights to tool-use LLM capabilities. Datasets and evaluation scripts of ToolSandbox are released at <placeholder>.

Synthetic data has recently proven effective in enhancing the accuracy of Text-to-SQL parsers. However, existing methods generate SQL queries first by randomly sampling tables and columns based on probability and then synthesize natural language questions (NLQs). This approach often produces a large number of NLQ-SQL pairs that are irrelevant to the target domain and inconsistent in query intent, significantly diminishing the fine-tuning effectiveness of LLMs. In this paper, we introduce DSQG-Syn, a novel text-to-SQL data synthesis framework that based on domain-specific question generation. Specifically, we design a question generation method that creates domain-relevant questions based on predefined question types, ensuring coverage of major SQL operations. Guided by these questions, we synthesize NLQ-SQL pairs that are both domain-relevant and intent-consistent. To further enhance data quality, we filter out noisy samples from the generated pairs. When popular open-source LLMs are fine-tuned on our high-quality synthesized dataset, they achieve significant accuracy improvements, surpassing the performance of closed-source LLM-based approaches. Moreover, we demonstrate that our method outperforms existing state-of-the-art (SOTA) data synthesis techniques.

Recent advances in large language models (LLMs) have increased the demand for comprehensive benchmarks to evaluate their capabilities as human-like agents. Existing benchmarks, while useful, often focus on specific application scenarios, emphasizing task completion but failing to dissect the underlying skills that drive these outcomes. This lack of granularity makes it difficult to deeply discern where failures stem from. Additionally, setting up these environments requires considerable effort, and issues of unreliability and reproducibility sometimes arise, especially in interactive tasks. To address these limitations, we introduce the Massive Multitask Agent Understanding (MMAU) benchmark, featuring comprehensive offline tasks that eliminate the need for complex environment setups. It evaluate models across five domains, including Tool-use, Directed Acyclic Graph (DAG) QA, Data Science and Machine Learning coding, Contest-level programming and Mathematics, and covering five essential capabilities: Understanding, Reasoning, Planning, Problem-solving, and Self-correction. With a total of 20 meticulously designed tasks encompassing over 3K distinct prompts, MMAU provides a comprehensive framework for evaluating the strengths and limitations of LLM agents. By testing 20 representative models on MMAU, we provide deep and insightful analyses. Ultimately, MMAU not only sheds light on the capabilities and limitations of LLM agents but also enhances the interpretability of their performance.

Current natural language processing models work well on a single task, yet they often fail to continuously learn new tasks without forgetting previous ones as they are re-trained throughout their lifetime, a challenge known as lifelong learning. State-of-the-art lifelong language learning methods store past examples in episodic memory and replay them at both training and inference time. However, as we show later in our experiments, there are three significant impediments: (1) needing unrealistically large memory module to achieve good performance, (2) suffering from negative transfer, (3) requiring multiple local adaptation steps for each test example that significantly slows down the inference speed. In this paper, we identify three common principles of lifelong learning methods and propose an efficient meta-lifelong framework that combines them in a synergistic fashion. To achieve sample efficiency, our method trains the model in a manner that it learns a better initialization for local adaptation. Extensive experiments on text classification and question answering benchmarks demonstrate the effectiveness of our framework by achieving state-of-the-art performance using merely 1% memory size and narrowing the gap with multi-task learning. We further show that our method alleviates both catastrophic forgetting and negative transfer at the same time.

Modern multilingual models are trained on concatenated text from multiple languages in hopes of conferring benefits to each (positive transfer), with the most pronounced benefits accruing to low-resource languages. However, recent work has shown that this approach can degrade performance on high-resource languages, a phenomenon known as negative interference. In this paper, we present the first systematic study of negative interference. We show that, contrary to previous belief, negative interference also impacts low-resource languages. While parameters are maximally shared to learn language-universal structures, we demonstrate that language-specific parameters do exist in multilingual models and they are a potential cause of negative interference. Motivated by these observations, we also present a meta-learning algorithm that obtains better cross-lingual transferability and alleviates negative interference, by adding language-specific layers as meta-parameters and training them in a manner that explicitly improves shared layers’ generalization on all languages. Overall, our results show that negative interference is more common than previously known, suggesting new directions for improving multilingual representations.