Language models have recently advanced into the realm of reasoning, yet it is through multimodal reasoning that we can fully unlock the potential to achieve more comprehensive, human-like cognitive capabilities. This survey provides a systematic overview of the recent multimodal reasoning approaches, categorizing them into two levels: language-centric multimodal reasoning and collaborative multimodal reasoning. The former encompasses one-pass visual perception and active visual perception, where vision primarily serves a supporting role in language reasoning. The latter involves action generation and state update within reasoning process, enabling a more dynamic interaction between modalities. Furthermore, we analyze the technical evolution of these methods, discuss their inherent challenges, and introduce key benchmark tasks and evaluation metrics for assessing multimodal reasoning performance. Finally, we provide insights into future research directions from the following two perspectives: (i) from visual-language reasoning to omnimodal reasoning and (ii) from multimodal reasoning to multimodal agents. This survey aims to provide a structured overview that will inspire further advancements in multimodal reasoning research.

The rapid development of large language models (LLMs) has brought significant attention to speech models, particularly recent progress in speech2speech protocols supporting speech input and output. However, the existing benchmarks adopt automatic text-based evaluators for evaluating the instruction following ability of these models lack consideration for paralinguistic information in both speech understanding and generation. To address these issues, we introduce S2S-Arena, a novel arena-style S2S benchmark that evaluates instruction-following capabilities with paralinguistic information in both speech-in and speech-out across real-world tasks. We design 154 samples that fused TTS and live recordings in four domains with 21 tasks and manually evaluate existing popular speech models in an arena-style manner. The experimental results show that: (1) in addition to the superior performance of GPT-4o, the speech model of cascaded ASR, LLM, and TTS outperforms the jointly trained model after text-speech alignment in speech2speech protocols; (2) considering paralinguistic information, the knowledgeability of the speech model mainly depends on the LLM backbone, and the multilingual support of that is limited by the speech module; (3) excellent speech models can already understand the paralinguistic information in speech input, but generating appropriate audio with paralinguistic information is still a challenge.

Pre-trained vision-language models (VLMs), such as CLIP, have exhibited remarkable performance across various downstream tasks by aligning text and images in a unified embedding space. However, due to the imbalanced distribution of pre-trained datasets, CLIP suffers from the bias problem in real-world applications. Existing debiasing methods struggle to obtain sufficient image samples for minority groups and incur high costs for group labeling. To address the limitations, we propose a Text-Only Debiasing framework called TOD, leveraging a text-as-image training paradigm to mitigate visual biases. Specifically, this approach repurposes the text encoder to function as an image encoder, thereby eliminating the need for image data. Simultaneously, it utilizes a large language model (LLM) to generate a balanced text dataset, which is then used for prompt tuning. However, we observed that the model overfits to the text modality because label names, serving as supervision signals, appear explicitly in the texts. To address this issue, we further introduce a Multi-Target Prediction (MTP) task that motivates the model to focus on complex contexts and distinguish between target and biased information. Extensive experiments on the Waterbirds and CelebA datasets show that our method significantly improves group robustness, achieving state-of-the-art results among image-free methods and even competitive performance compared to image-supervised methods. Furthermore, the proposed method can be adapted to challenging scenarios with multiple or unknown bias attributes, demonstrating its strong generalization and robustness.

Generative Artificial Intelligence (AI) encounters limitations in efficiency and fairness within the realm of Procedural Content Generation (PCG) when human creators solely drive and bear responsibility for the generative process. Alternative setups, such as Mixed-Initiative Co-Creative (MI-CC) systems, exhibited their promise. Still, the potential of an active mixed initiative, where AI takes a role beyond following, is understudied. This work investigates the influence of the adaptive ability of an active and learning AI agent on creators' expectancy of creative responsibilities in an MI-CC setting. We built and studied a system that employs reinforcement learning (RL) methods to learn the creative responsibility preferences of a human user during online interactions. Situated in story co-creation, we develop a Multi-armed-bandit agent that learns from the human creator, updates its collaborative decision-making belief, and switches between its capabilities during an MI-CC experience. With 39 participants joining a human subject study, Our developed system's learning capabilities are well recognized compared to the non-learning ablation, corresponding to a significant increase in overall satisfaction with the MI-CC experience. These findings indicate a robust association between effective MI-CC collaborative interactions, particularly the implementation of proactive AI initiatives, and deepened understanding among all participants.

Unlike reinforcement learning (RL) agents, humans remain capable multitaskers in changing environments. In spite of only experiencing the world through their own observations and interactions, people know how to balance focusing on tasks with learning about how changes may affect their understanding of the world. This is possible by choosing to solve tasks in ways that are interesting and generally informative beyond just the current task. Motivated by this, we propose an agent influence framework for RL agents to improve the adaptation efficiency of external models in changing environments without any changes to the agent's rewards. Our formulation is composed of two self-contained modules: interest fields and behavior shaping via interest fields. We implement an uncertainty-based interest field algorithm as well as a skill-sampling-based behavior-shaping algorithm to use in testing this framework. Our results show that our method outperforms the baselines in terms of external model adaptation on metrics that measure both efficiency and performance.

The term co-creativity has been used to describe a wide variety of human-AI assemblages in which human and AI are both involved in a creative endeavor. In order to assist with disambiguating research efforts, we present an ontology of co-creative systems, focusing on how responsibilities are divided between human and AI system and the information exchanged between them. We extend Lubart's original ontology of creativity support tools with three new categories emphasizing artificial intelligence: computer-as-subcontractor, computer-as-critic, and computer-as-teammate, some of which have sub-categorizations.

Many advancements have been made in procedural content generation for games, and with mixed-initiative co-creativity, have the potential for great benefits to human designers. However, co-creative systems for game generation are typically limited to specific genres, rules, or games, limiting the creativity of the designer. We seek to model games abstractly enough to apply to any genre, focusing on designing game systems and mechanics, and create a controllable, co-creative agent that can collaborate on these designs. We present a model of games using state-machine-like components and resource flows, a set of controllable metrics, a design evaluator simulating playthroughs with these metrics, and an evolutionary design balancer and generator. We find this system to be both able to express a wide range of games and able to be human-controllable for future co-creative applications.

Generative Artificial Intelligence systems have been developed for image, code, story, and game generation with the goal of facilitating human creativity. Recent work on neural generative systems has emphasized one particular means of interacting with AI systems: the user provides a specification, usually in the form of prompts, and the AI system generates the content. However, there are other configurations of human and AI coordination, such as co-creativity (CC) in which both human and AI systems can contribute to content creation, and mixed-initiative (MI) in which both human and AI systems can initiate content changes. In this paper, we define a hypothetical human-AI configuration design space consisting of different means for humans and AI systems to communicate creative intent to each other. We conduct a human participant study with 185 participants to understand how users want to interact with differently configured MI-CC systems. We find out that MI-CC systems with more extensive coverage of the design space are rated higher or on par on a variety of creative and goal-completion metrics, demonstrating that wider coverage of the design space can improve user experience and achievement when using the system; Preference varies greatly between expertise groups, suggesting the development of adaptive, personalized MI-CC systems; Participants identified new design space dimensions including scrutability -- the ability to poke and prod at models -- and explainability.

Open-world novelty--a sudden change in the mechanics or properties of an environment--is a common occurrence in the real world. Novelty adaptation is an agent's ability to improve its policy performance post-novelty. Most reinforcement learning (RL) methods assume that the world is a closed, fixed process. Consequentially, RL policies adapt inefficiently to novelties. To address this, we introduce WorldCloner, an end-to-end trainable neuro-symbolic world model for rapid novelty adaptation. WorldCloner learns an efficient symbolic representation of the pre-novelty environment transitions, and uses this transition model to detect novelty and efficiently adapt to novelty in a single-shot fashion. Additionally, WorldCloner augments the policy learning process using imagination-based adaptation, where the world model simulates transitions of the post-novelty environment to help the policy adapt. By blending ''imagined'' transitions with interactions in the post-novelty environment, performance can be recovered with fewer total environment interactions. Using environments designed for studying novelty in sequential decision-making problems, we show that the symbolic world model helps its neural policy adapt more efficiently than model-based and model-based neural-only reinforcement learning methods.

The security of artificial intelligence (AI) is an important research area towards safe, reliable, and trustworthy AI systems. To accelerate the research on AI security, the Artificial Intelligence Security Competition (AISC) was organized by the Zhongguancun Laboratory, China Industrial Control Systems Cyber Emergency Response Team, Institute for Artificial Intelligence, Tsinghua University, and RealAI as part of the Zhongguancun International Frontier Technology Innovation Competition (https://www.zgc-aisc.com/en). The competition consists of three tracks, including Deepfake Security Competition, Autonomous Driving Security Competition, and Face Recognition Security Competition. This report will introduce the competition rules of these three tracks and the solutions of top-ranking teams in each track.