The recent surge in Multimodal Large Language Models (MLLMs) has showcased their remarkable potential for achieving generalized intelligence by integrating visual understanding into Large Language Models. Nevertheless, the sheer model size of MLLMs leads to substantial memory and computational demands that hinder their widespread deployment. In this work, we do not propose a new efficient model structure or train small-scale MLLMs from scratch. Instead, we focus on what matters for training small-scale MLLMs through knowledge distillation, which is the first step from the multimodal distillation perspective. Our extensive studies involve training strategies, model choices, and distillation algorithms in the knowledge distillation process. These results show that joint alignment for both tokens and logit alignment plays critical roles in teacher-student frameworks. In addition, we draw a series of intriguing observations from this study. By evaluating different benchmarks and proper strategy, even a 2.7B small-scale model can perform on par with larger models with 7B or 13B parameters. Our code and models will be publicly available for further research.

Developing text-based robot trajectory generation models is made particularly difficult by the small dataset size, high dimensionality of the trajectory space, and the inherent complexity of the text-conditional motion distribution. Recent manifold learning-based methods have partially addressed the dimensionality and dataset size issues, but struggle with the complex text-conditional distribution. In this paper we propose a text-based trajectory generation model that attempts to address all three challenges while relying on only a handful of demonstration trajectory data. Our key idea is to leverage recent flow-based models capable of capturing complex conditional distributions, not directly in the high-dimensional trajectory space, but rather in the low-dimensional latent coordinate space of the motion manifold, with deliberately designed regularization terms to ensure smoothness of motions and robustness to text variations. We show that our {\it Motion Manifold Flow Primitive (MMFP)} framework can accurately generate qualitatively distinct motions for a wide range of text inputs, significantly outperforming existing methods.

Addressing school bullying effectively and promptly is crucial for the mental health of students. This study examined the potential of large language models (LLMs) to empower students by discerning between bullying and joking in school peer interactions. We employed ChatGPT-4, Gemini 1.5 Pro, and Claude 3 Opus, evaluating their effectiveness through human review. Our results revealed that not all LLMs were suitable for an agentic approach, with ChatGPT-4 showing the most promise. We observed variations in LLM outputs, possibly influenced by political overcorrectness, context window limitations, and pre-existing bias in their training data. ChatGPT-4 excelled in context-specific accuracy after implementing the agentic approach, highlighting its potential to provide continuous, real-time support to vulnerable students.

Enhancing interaction and feedback collection in a first-semester computer science course poses a significant challenge due to students' diverse needs and engagement levels. To address this issue, we created and integrated a command-based chatbot on the course communication server on Discord. The DiscordBot enables students to provide feedback on course activities through short surveys, such as exercises, quizzes, and lectures, facilitating stress-free communication with instructors. It also supports attendance tracking and introduces lectures before they start. The research demonstrates the effectiveness of the DiscordBot as a communication tool. The ongoing feedback allowed course instructors to dynamically adjust and improve the difficulty level of upcoming activities and promote discussion in subsequent tutor sessions. The data collected reveal that students can accurately perceive the activities' difficulty and expected results, providing insights not possible through traditional end-of-semester surveys. Students reported that interaction with the DiscordBot was easy and expressed a desire to continue using it in future semesters. This responsive approach ensures the course meets the evolving needs of students, thereby enhancing their overall learning experience.

Humans appear to have a critical period (CP) for language acquisition: Second language (L2) acquisition becomes harder after early childhood, and ceasing exposure to a first language (L1) after this period (but not before) typically does not lead to substantial loss of L1 proficiency. It is unknown whether these CP effects result from innately determined brain maturation or as a stabilization of neural connections naturally induced by experience. In this study, we use language models (LMs) to test the extent to which these phenomena are peculiar to humans, or shared by a broader class of language learners. We vary the age of exposure by training LMs on language pairs in various experimental conditions, and find that LMs, which lack any direct analog to innate maturational stages, do not show CP effects when trained sequentially on L1 and L2. Our results contradict the claim that CP effects are an inevitable result of learning in statistical learners, and they are consistent with an innate mechanism for CP effects. We show that we can reverse-engineer the CP by introducing a regularizer partway through training to simulate a maturational decrease in plasticity. All in all, our results suggest that L1 learning on its own may not be enough to induce a CP, and additional engineering is necessary to make language models more cognitively plausible.

AI models can easily generate essays and other types of text. However, they're nowhere near as good at solving math problems, which tend to involve logical reasoning--something that's beyond the capabilities of most current AI systems. But that may finally be changing. Google DeepMind says it has trained two specialized AI systems to solve complex math problems involving advanced reasoning. The systems worked together to successfully solve four out of six problems from this year's International Mathematical Olympiad, a prestigious competition for high school students.

Recent advancements have significantly enhanced the capabilities of Multimodal Large Language Models (MLLMs) in generating and understanding image-to-text content. Despite these successes, progress is predominantly limited to English due to the scarcity of high quality multimodal resources in other languages. This limitation impedes the development of competitive models in languages such as Arabic. To alleviate this situation, we introduce an efficient Arabic multimodal assistant, dubbed Dallah, that utilizes an advanced language model based on LLaMA-2 to facilitate multimodal interactions. Dallah demonstrates state-of-the-art performance in Arabic MLLMs. Through fine-tuning six Arabic dialects, Dallah showcases its capability to handle complex dialectal interactions incorporating both textual and visual elements. The model excels in two benchmark tests: one evaluating its performance on Modern Standard Arabic (MSA) and another specifically designed to assess dialectal responses. Beyond its robust performance in multimodal interaction tasks, Dallah has the potential to pave the way for further development of dialect-aware Arabic MLLMs.

Calls for engagement with the public in Artificial Intelligence (AI) research, development, and governance are increasing, leading to the use of surveys to capture people's values, perceptions, and experiences related to AI. In this paper, we critically examine the state of human participant surveys associated with these topics. Through both a reflexive analysis of a survey pilot spanning six countries and a systematic literature review of 44 papers featuring public surveys related to AI, we explore prominent perspectives and methodological nuances associated with surveys to date. We find that public surveys on AI topics are vulnerable to specific Western knowledge, values, and assumptions in their design, including in their positioning of ethical concepts and societal values, lack sufficient critical discourse surrounding deployment strategies, and demonstrate inconsistent forms of transparency in their reporting. Based on our findings, we distill provocations and heuristic questions for our community, to recognize the limitations of surveys for meeting the goals of engagement, and to cultivate shared principles to design, deploy, and interpret surveys cautiously and responsibly.

A central piece in enabling intelligent agentic behavior in foundation models is to make them capable of introspecting upon their behavior, reasoning, and correcting their mistakes as more computation or interaction is available. Even the strongest proprietary large language models (LLMs) do not quite exhibit the ability of continually improving their responses sequentially, even in scenarios where they are explicitly told that they are making a mistake. In this paper, we develop RISE: Recursive IntroSpEction, an approach for fine-tuning LLMs to introduce this capability, despite prior work hypothesizing that this capability may not be possible to attain. Our approach prescribes an iterative fine-tuning procedure, which attempts to teach the model how to alter its response after having executed previously unsuccessful attempts to solve a hard test-time problem, with optionally additional environment feedback. RISE poses fine-tuning for a single-turn prompt as solving a multi-turn Markov decision process (MDP), where the initial state is the prompt. Inspired by principles in online imitation learning and reinforcement learning, we propose strategies for multi-turn data collection and training so as to imbue an LLM with the capability to recursively detect and correct its previous mistakes in subsequent iterations. Our experiments show that RISE enables Llama2, Llama3, and Mistral models to improve themselves with more turns on math reasoning tasks, outperforming several single-turn strategies given an equal amount of inference-time computation. We also find that RISE scales well, often attaining larger benefits with more capable models. Our analysis shows that RISE makes meaningful improvements to responses to arrive at the correct solution for challenging prompts, without disrupting one-turn abilities as a result of expressing more complex distributions.

Training a semi-supervised end-to-end speech recognition system using noisy student training has significantly improved performance. However, this approach requires a substantial amount of paired speech-text and unlabeled speech, which is costly for low-resource languages. Therefore, this paper considers a more extreme case of semi-supervised end-to-end automatic speech recognition where there are limited paired speech-text, unlabeled speech (less than five hours), and abundant external text. Firstly, we observe improved performance by training the model using our previous work on semi-supervised learning "CycleGAN and inter-domain losses" solely with external text. Secondly, we enhance "CycleGAN and inter-domain losses" by incorporating automatic hyperparameter tuning, calling it "enhanced CycleGAN inter-domain losses." Thirdly, we integrate it into the noisy student training approach pipeline for low-resource scenarios. Our experimental results, conducted on six non-English languages from Voxforge and Common Voice, show a 20% word error rate reduction compared to the baseline teacher model and a 10% word error rate reduction compared to the baseline best student model, highlighting the significant improvements achieved through our proposed method.