The field of multimodal reasoning and large language models is rapidly advancing, with a focus on improving the efficiency, accuracy, and adaptability of these models. Recent developments have seen the introduction of new architectures, such as the Mixture-of-Experts (MoE) model, which enables more efficient and scalable processing of large amounts of data. Additionally, there has been a growing interest in multimodal reasoning, with models being developed to handle multiple forms of input, such as text, images, and speech. Noteworthy papers in this area include the proposal of pre-attention expert prediction for MoE models, which achieves state-of-the-art results on several benchmarks, and the introduction of the PRISM framework for user-centric conversational stance detection, which demonstrates significant gains over strong baselines. Other notable papers include the development of the Uni-MoE-2.0-Omni model, which achieves highly competitive performance on a wide range of benchmarks, and the proposal of the MMD-Thinker framework for multimodal misinformation detection, which achieves state-of-the-art performance on several datasets.
Advances in Multimodal Reasoning and Large Language Models
Sources
Pre-Attention Expert Prediction and Prefetching for Mixture-of-Experts Large Language Models
PRISM of Opinions: A Persona-Reasoned Multimodal Framework for User-centric Conversational Stance Detection
More Than Irrational: Modeling Belief-Biased Agents
Learning to Trust: Bayesian Adaptation to Varying Suggester Reliability in Sequential Decision Making
Uni-MoE-2.0-Omni: Scaling Language-Centric Omnimodal Large Model with Advanced MoE, Training and Data
Optimal Foraging in Memory Retrieval: Evaluating Random Walks and Metropolis-Hastings Sampling in Modern Semantic Spaces
MMD-Thinker: Adaptive Multi-Dimensional Thinking for Multimodal Misinformation Detection
HiEAG: Evidence-Augmented Generation for Out-of-Context Misinformation Detection
Ask WhAI:Probing Belief Formation in Role-Primed LLM Agents
Octopus: Agentic Multimodal Reasoning with Six-Capability Orchestration
MoDES: Accelerating Mixture-of-Experts Multimodal Large Language Models via Dynamic Expert Skipping
ChemLabs on ChemO: A Multi-Agent System for Multimodal Reasoning on IChO 2025
OpenMMReasoner: Pushing the Frontiers for Multimodal Reasoning with an Open and General Recipe
Robot Metacognition: Decision Making with Confidence for Tool Invention
EvoLMM: Self-Evolving Large Multimodal Models with Continuous Rewards