Self-correction is a novel method that can stimulate the potential reasoning abilities of large language models (LLMs). It involves detecting and correcting errors during the inference process when LLMs solve reasoning problems. However, recent works do not regard self-correction as a spontaneous and intrinsic capability of LLMs. Instead, such correction is achieved through post-hoc generation, external knowledge introduction, multi-model collaboration, and similar techniques. In this paper, we propose a series of mathematical LLMs called S$^3$c-Math, which are able to perform Spontaneous Step-level Self-correction for Mathematical reasoning. This capability helps LLMs to recognize whether their ongoing inference tends to contain errors and simultaneously correct these errors to produce a more reliable response. We proposed a method, which employs a step-level sampling approach to construct step-wise self-correction data for achieving such ability. Additionally, we implement a training strategy that uses above constructed data to equip LLMs with spontaneous step-level self-correction capacities. Our data and methods have been demonstrated to be effective across various foundation LLMs, consistently showing significant progress in evaluations on GSM8K, MATH, and other mathematical benchmarks. To the best of our knowledge, we are the first to introduce the spontaneous step-level self-correction ability of LLMs in mathematical reasoning.

Current LLM training positions mathematical reasoning as a core capability. With publicly available sources fully tapped, there is unmet demand for diverse and challenging math questions. Relying solely on human experts is both time-consuming and costly, while LLM-generated questions often lack the requisite diversity and difficulty. We present a design framework that combines the strengths of LLMs with a human-in-the-loop approach to generate a diverse array of challenging math questions. We leverage LLM metacognition skills [Didolkar et al., 2024] of a strong LLM to extract core "skills" from existing math datasets. These skills serve as the basis for generating novel and difficult questions by prompting the LLM with random pairs of core skills. The use of two different skills within each question makes finding such questions an "out of distribution" task for both LLMs and humans. Our pipeline employs LLMs to iteratively generate and refine questions and solutions through multiturn prompting. Human annotators then verify and further refine the questions, with their efficiency enhanced via further LLM interactions. Applying this pipeline on skills extracted from the MATH dataset [Hendrycks et al., 2021] resulted in MATH$^2$ - a dataset of higher-quality math questions, as evidenced by: (a) Lower performance of all models on MATH$^2$ than on MATH (b) Higher performance on MATH when using MATH$^2$ questions as in-context examples. Although focused on mathematics, our methodology seems applicable to other domains requiring structured reasoning, and potentially as a component of scalable oversight. Also of interest is a striking relationship observed between models' performance on the new dataset: the success rate on MATH$^2$ is the square on MATH, suggesting that successfully solving the question in MATH$^2$ requires a nontrivial combination of two distinct math skills.

Sign languages, used by around 70 million Deaf individuals globally, are visual languages that convey visual and contextual information. Current methods in vision-based sign language recognition (SLR) and translation (SLT) struggle with dialogue scenes due to limited dataset diversity and the neglect of contextually relevant information. To address these challenges, we introduce SCOPE (Sign language Contextual Processing with Embedding from LLMs), a novel context-aware vision-based SLR and SLT framework. For SLR, we utilize dialogue contexts through a multi-modal encoder to enhance gloss-level recognition. For subsequent SLT, we further fine-tune a Large Language Model (LLM) by incorporating prior conversational context. We also contribute a new sign language dataset that contains 72 hours of Chinese sign language videos in contextual dialogues across various scenarios. Experimental results demonstrate that our SCOPE framework achieves state-of-the-art performance on multiple datasets, including Phoenix-2014T, CSL-Daily, and our SCOPE dataset. Moreover, surveys conducted with participants from the Deaf community further validate the robustness and effectiveness of our approach in real-world applications. Both our dataset and code will be open-sourced to facilitate further research.

Beijing Institute for Wangxuan Institute of Computer Technology, Peking University General Artificial Intelligence National Key Laboratory of General Artificial Intelligence wangyuxuan1@bigai.ai Hallucination is a common issue in Multimodal Large Language Models (MLLMs), yet the underlying principles remain poorly understood. In this paper, we investigate which components of MLLMs contribute to object hallucinations. To analyze image representations while completely avoiding the influence of all other factors other than the image representation itself, we propose a parametricfree representation alignment metric (Pfram) that can measure the similarities between any two representation systems without requiring additional training parameters. Notably, Pfram can also assess the alignment of a neural representation system with the human representation system, represented by ground-truth annotations of images. By evaluating the alignment with object annotations, we demonstrate that this metric shows strong and consistent correlations with object hallucination across a wide range of state-of-the-art MLLMs, spanning various model architectures and sizes. Furthermore, using this metric, we explore other key issues related to image representations in MLLMs, such as the role of different modules, the impact of textual instructions, and potential improvements including the use of alternative visual encoders. Multimodal Large Language Models (MLLMs) have been rapidly advancing in recent days Dai et al. (2023); Liu et al. (2023c;b); Zhang et al. (2023); Dong et al. (2024); Bai et al. (2023).

Although EnCLAP exhibits impressive performance, the study by Kim et al. lacks sufficient experimental evaluation for determining In this work, we aim to analyze and optimize the EnCLAP framework, the optimal models for the model components. Notably, a state-of-the-art model in automated audio captioning. We the authors do not investigate alternative sequence-level acoustic investigate the impact of modifying the acoustic encoder components, features beyond CLAP. Furthermore, for timestep-level acoustic explore pretraining with different dataset scales, and study the features, while they demonstrate that discrete codec input outperforms effectiveness of a reranking scheme. Through extensive experimentation continuous input, their analysis is restricted to a single setup and quantitative analysis of generated captions, we develop using EnCodec, without exploring other options or configurations. EnCLAP++, an enhanced version that significantly surpasses the Additionally, Kim et al. acknowledge the issue of overfitting in original.

Sensitive information detection is crucial in content moderation to maintain safe online communities. Assisting in this traditionally manual process could relieve human moderators from overwhelming and tedious tasks, allowing them to focus solely on flagged content that may pose potential risks. Rapidly advancing large language models (LLMs) are known for their capability to understand and process natural language and so present a potential solution to support this process. This study explores the capabilities of five LLMs for detecting sensitive messages in the mental well-being domain within two online datasets and assesses their performance in terms of accuracy, precision, recall, F1 scores, and consistency. Our findings indicate that LLMs have the potential to be integrated into the moderation workflow as a convenient and precise detection tool. The best-performing model, GPT-4o, achieved an average accuracy of 99.5\% and an F1-score of 0.99. We discuss the advantages and potential challenges of using LLMs in the moderation workflow and suggest that future research should address the ethical considerations of utilising this technology.

Background: This study aimed to evaluate and compare the performance of classical machine learning models (CMLs) and large language models (LLMs) in predicting mortality associated with COVID-19 by utilizing a high-dimensional tabular dataset. Materials and Methods: We analyzed data from 9,134 COVID-19 patients collected across four hospitals. Seven CML models, including XGBoost and random forest (RF), were trained and evaluated. The structured data was converted into text for zero-shot classification by eight LLMs, including GPT-4 and Mistral-7b. Additionally, Mistral-7b was fine-tuned using the QLoRA approach to enhance its predictive capabilities. Results: Among the CML models, XGBoost and RF achieved the highest accuracy, with F1 scores of 0.87 for internal validation and 0.83 for external validation. In the LLM category, GPT-4 was the top performer with an F1 score of 0.43. Fine-tuning Mistral-7b significantly improved its recall from 1% to 79%, resulting in an F1 score of 0.74, which was stable during external validation. Conclusion: While LLMs show moderate performance in zero-shot classification, fine-tuning can significantly enhance their effectiveness, potentially aligning them closer to CML models. However, CMLs still outperform LLMs in high-dimensional tabular data tasks.

Monocular depth estimation is a critical function in computer vision applications. This paper shows that large language models (LLMs) can effectively interpret depth with minimal supervision, using efficient resource utilization and a consistent neural network architecture. We introduce LLM-MDE, a multimodal framework that deciphers depth through language comprehension. Specifically, LLM-MDE employs two main strategies to enhance the pretrained LLM's capability for depth estimation: cross-modal reprogramming and an adaptive prompt estimation module. These strategies align vision representations with text prototypes and automatically generate prompts based on monocular images, respectively. Comprehensive experiments on real-world MDE datasets confirm the effectiveness and superiority of LLM-MDE, which excels in few-/zero-shot tasks while minimizing resource use. The source code is available.

Metaphor and sarcasm are common figurative expressions in people's communication, especially on the Internet or the memes popular among teenagers. We create a new benchmark named NYK-MS (NewYorKer for Metaphor and Sarcasm), which contains 1,583 samples for metaphor understanding tasks and 1,578 samples for sarcasm understanding tasks. These tasks include whether it contains metaphor/sarcasm, which word or object contains metaphor/sarcasm, what does it satirize and why does it contains metaphor/sarcasm, all of the 7 tasks are well-annotated by at least 3 annotators. We annotate the dataset for several rounds to improve the consistency and quality, and use GUI and GPT-4V to raise our efficiency. Based on the benchmark, we conduct plenty of experiments. In the zero-shot experiments, we show that Large Language Models (LLM) and Large Multi-modal Models (LMM) can't do classification task well, and as the scale increases, the performance on other 5 tasks improves. In the experiments on traditional pre-train models, we show the enhancement with augment and alignment methods, which prove our benchmark is consistent with previous dataset and requires the model to understand both of the two modalities.

Machine learning research, crucial for technological advancements and innovation, often faces significant challenges due to its inherent complexity, slow pace of experimentation, and the necessity for specialized expertise. Motivated by this, we present a new systematic framework, autonomous Machine Learning Research with large language models (MLR-Copilot), designed to enhance machine learning research productivity through the automatic generation and implementation of research ideas using Large Language Model (LLM) agents. The framework consists of three phases: research idea generation, experiment implementation, and implementation execution. First, existing research papers are used to generate hypotheses and experimental plans vis IdeaAgent powered by LLMs. Next, the implementation generation phase translates these plans into executables with ExperimentAgent. This phase leverages retrieved prototype code and optionally retrieves candidate models and data. Finally, the execution phase, also managed by ExperimentAgent, involves running experiments with mechanisms for human feedback and iterative debugging to enhance the likelihood of achieving executable research outcomes. We evaluate our framework on five machine learning research tasks and the experimental results show the framework's potential to facilitate the research progress and innovations.