Solving complex chart Q&A tasks requires advanced visual reasoning abilities in multimodal large language models (MLLMs). Recent studies highlight that these abilities consist of two main parts: recognizing key information from visual inputs and conducting reasoning over it. Thus, a promising approach to enhance MLLMs is to construct relevant training data focusing on the two aspects. However, collecting and annotating complex charts and questions is costly and timeconsuming, and ensuring the quality of annotated answers remains a challenge. In this paper, we propose Code-as-Intermediary Translation (CIT), a cost-effective, efficient and easily scalable data synthesis method for distilling visual reasoning abilities from LLMs to MLLMs. The code serves as an intermediary that translates visual chart representations into textual representations, enabling LLMs to understand cross-modal information. QA, a dataset containing 3k reasoning-intensive charts and 20k Q&A pairs to enhance both recognition and reasoning abilities. Experiments show that when fine-tuned with our data, models not only perform well on chart-related benchmarks, but also demonstrate improved multimodal reasoning abilities on general mathematical benchmarks like MathVista. Multimodal large language models (MLLMs) have made significant achievements, particularly in visual recognition tasks (OpenAI, 2024a; Anthropic, 2024). While they can handle simple visual inputs well, there has been a growing emphasis on complex chart understanding, driven by the widespread use of charts in real-world contexts (Masry et al., 2022; Huang et al., 2024). However, addressing reasoning-intensive questions involving charts remains challenging for these models. Existing benchmarks underscore the need for more advanced and generalized visual reasoning abilities, which are still underdeveloped in current MLLMs (Wang et al., 2024c; Lu et al., 2024). Our analysis of the error distribution in ChartQA (Figure 1) also highlights two main types of model failure: 62% of errors stem from misrecognition, while 36% arise from reasoning mistakes after correct recognition. This shows that even advanced MLLMs struggle with basic recognition and often make superficial reasoning errors. In contrast, humans excel at these tasks by purposefully identifying query-relevant information from images and engaging in step-by-step reasoning (Wang et al., 2024c;a). In light of these findings, enabling models to solve problems in a human-like manner, becomes essential for advancing visual reasoning performance. One promising strategy is to distill the rationales of reasoning from experts, such as human or stronger models (Han et al., 2023; Meng et al., 2024; Masry et al., 2024a;b) However, creating highquality training data for chart-related tasks is costly and time-consuming.

Large language models (LLMs) and large multimodal models (LMMs) have shown great potential in automating complex tasks like web browsing and gaming. However, their ability to generalize across diverse applications remains limited, hindering broader utility. To address this challenge, we present OSCAR: Operating System Control via state-Aware reasoning and Re-planning. OSCAR is a generalist agent designed to autonomously navigate and interact with various desktop and mobile applications through standardized controls, such as mouse and keyboard inputs, while processing screen images to fulfill user commands. To enhance stability and adaptability, OSCAR operates as a state machine, equipped with error-handling mechanisms and task-driven re-planning, allowing it to efficiently adjust to real-time feedback and exceptions. We demonstrate OSCAR's effectiveness through extensive experiments on diverse benchmarks across desktop and mobile platforms, where it transforms complex workflows into simple natural language commands, significantly boosting user productivity. Our code will be open-source upon publication. These model-centric agents show revolutionary potential for automating real-world tasks such as web browsing (Gur et al., 2023), gaming (Krzywinska, 2024), and software development (Hong et al.). However, despite impressive results, these agents struggle to generalize across different applications due to variations in observation and action spaces. In real-world scenarios, workflows often involve switching between applications and interacting with diverse graphical or command-line interfaces. This raises an intriguing and practical question: can we build a generalist agent capable of following user instructions across various applications using standardized operating system (OS) controls like mouse and keyboard inputs, while processing screen outputs?

Recent research has focused on literary machine translation (MT) as a new challenge in MT. However, the evaluation of literary MT remains an open problem. We contribute to this ongoing discussion by introducing LITEVAL-CORPUS, a paragraph-level parallel corpus comprising multiple verified human translations and outputs from 9 MT systems, which totals over 2k paragraphs and includes 13k annotated sentences across four language pairs, costing 4.5k Euro. This corpus enables us to (i) examine the consistency and adequacy of multiple annotation schemes, (ii) compare evaluations by students and professionals, and (iii) assess the effectiveness of LLM-based metrics. We find that Multidimensional Quality Metrics (MQM), as the de facto standard in non-literary human MT evaluation, is inadequate for literary translation: While Best-Worst Scaling (BWS) with students and Scalar Quality Metric (SQM) with professional translators prefer human translations at rates of ~82% and ~94%, respectively, MQM with student annotators prefers human professional translations over the translations of the best-performing LLMs in only ~42% of cases. While automatic metrics generally show a moderate correlation with human MQM and SQM, they struggle to accurately identify human translations, with rates of at most ~20%. Our overall evaluation indicates that human professional translations consistently outperform LLM translations, where even the most recent LLMs tend to produce more literal and less diverse translations compared to human translations. However, newer LLMs such as GPT-4o perform substantially better than older ones.

We introduce the Principled Reasoning and Acting (PRAct) framework, a novel method for learning and enforcing action principles from trajectory data. Central to our approach is the use of text gradients from a reflection and optimization engine to derive these action principles. To adapt action principles to specific task requirements, we propose a new optimization framework, Reflective Principle Optimization (RPO). After execution, RPO employs a reflector to critique current action principles and an optimizer to update them accordingly. We develop the RPO framework under two scenarios: Reward-RPO, which uses environmental rewards for reflection, and Self-RPO, which conducts self-reflection without external rewards. Additionally, two RPO methods, RPO-Traj and RPO-Batch, is introduced to adapt to different settings. Experimental results across four environments demonstrate that the PRAct agent, leveraging the RPO framework, effectively learns and applies action principles to enhance performance.

Tree of Thoughts (ToT) is a reasoning strategy for Large Language Models (LLMs) that employs a generator to suggest reasoning steps and a discriminator to decide which steps to implement. ToT demonstrates strong performance on reasoning tasks, often surpassing simple methods such as Input-Output (IO) prompting and Chain-of-Thought (CoT) reasoning. However, ToT does not consistently outperform such simpler methods across all models, leaving large knowledge gaps on the conditions under which ToT is most beneficial. In this paper, we analyze the roles of the generator and discriminator separately to better understand the conditions when ToT is beneficial. We find that the generator plays a more critical role than the discriminator in driving the success of ToT. Scaling the generator leads to notable improvements in ToT performance, even when using a smaller model as the discriminator, whereas scaling the discriminator with a fixed generator yields only marginal gains. Our results show that models across different scales exhibit comparable discrimination capabilities, yet differ significantly in their generative performance for ToT.

Large language models (LLMs) have taken the world by storm by making many previously difficult uses of AI feasible. LLMs are controlled via highly expressive textual prompts and return textual answers. Unfortunately, this unstructured text as input and output makes LLM-based applications brittle. This motivates the rise of prompting frameworks, which mediate between LLMs and the external world. However, existing prompting frameworks either have a high learning curve or take away control over the exact prompts from the developer. To overcome this dilemma, this paper introduces the Prompt Declaration Language (PDL). PDL is a simple declarative data-oriented language that puts prompts at the forefront, based on YAML. PDL works well with many LLM platforms and LLMs. It supports writing interactive applications that call LLMs and tools, and makes it easy to implement common use-cases such as chatbots, RAG, or agents. We hope PDL will make prompt programming simpler, less brittle, and more enjoyable.

Narratives are widely recognized as a powerful tool for structuring information and facilitating comprehension of complex ideas in various domains such as science communication. This paper investigates whether incorporating narrative elements can assist Large Language Models (LLMs) in solving complex problems more effectively. We propose a novel approach, Story of Thought (SoT), integrating narrative structures into prompting techniques for problem-solving. This approach involves constructing narratives around problem statements and creating a framework to identify and organize relevant information. Our experiments show that using various LLMs with SoT consistently surpasses using them with other techniques on physics, chemistry, math, and biology questions in both the GPQA and JEEBench datasets. The narrative-based information curation process in SoT enhances problem comprehension by contextualizing critical in-domain information and highlighting causal relationships within the problem space.

Academic paper search is an essential task for efficient literature discovery and scientific advancement. While dense retrieval has advanced various ad-hoc searches, it often struggles to match the underlying academic concepts between queries and documents, which is critical for paper search. To enable effective academic concept matching for paper search, we propose Taxonomy-guided Semantic Indexing (TaxoIndex) framework. TaxoIndex extracts key concepts from papers and organizes them as a semantic index guided by an academic taxonomy, and then leverages this index as foundational knowledge to identify academic concepts and link queries and documents. As a plug-and-play framework, TaxoIndex can be flexibly employed to enhance existing dense retrievers. Extensive experiments show that TaxoIndex brings significant improvements, even with highly limited training data, and greatly enhances interpretability.

Large language models (LLMs) such as ChatGPT are fine-tuned on large and diverse instruction-following corpora, and can generalize to new tasks. However, those instruction-tuned LLMs often perform poorly in specialized medical natural language understanding (NLU) tasks that require domain knowledge, granular text comprehension, and structured data extraction. To bridge the gap, we: (1) propose a unified prompting format for 7 important NLU tasks, % through span extraction and multi-choice question-answering (QA), (2) curate an instruction-tuning dataset, MNLU-Instruct, utilizing diverse existing open-source medical NLU corpora, and (3) develop BioMistral-NLU, a generalizable medical NLU model, through fine-tuning BioMistral on MNLU-Instruct. We evaluate BioMistral-NLU in a zero-shot setting, across 6 important NLU tasks, from two widely adopted medical NLU benchmarks: Biomedical Language Understanding Evaluation (BLUE) and Biomedical Language Understanding and Reasoning Benchmark (BLURB). Our experiments show that our BioMistral-NLU outperforms the original BioMistral, as well as the proprietary LLMs - ChatGPT and GPT-4. Our dataset-agnostic prompting strategy and instruction tuning step over diverse NLU tasks enhance LLMs' generalizability across diverse medical NLU tasks. Our ablation experiments show that instruction-tuning on a wider variety of tasks, even when the total number of training instances remains constant, enhances downstream zero-shot generalization.

Recent advances in large language models (LLMs) have shown promise for scalable educational applications, but their use in dialog-based tutoring systems remains challenging due to the need for effective pedagogical strategies and the high costs associated with expert-curated datasets. Our study explores the use of smaller, more affordable LLMs for one-on-one tutoring in the context of solving reading comprehension problems. We developed a synthetic tutoring dialog dataset, evaluated by human teachers, and fine-tuned a smaller LLM using this dataset. Furthermore, we conducted an interactive experiment comparing the performance of the fine-tuned model with a larger model in real-world tutoring scenarios. Our results show that the fine-tuned model performs on par with the larger model but at a lower cost, demonstrating a viable, cost-effective approach for implementing LLM-based tutoring systems in educational settings.