The proliferation of multimodal memes in the social media era demands that multimodal Large Language Models (mLLMs) effectively understand meme harmfulness. Existing benchmarks for assessing mLLMs on harmful meme understanding rely on accuracy-based, model-agnostic evaluations using static datasets. These benchmarks are limited in their ability to provide up-to-date and thorough assessments, as online memes evolve dynamically. To address this, we propose AdamMeme, a flexible, agent-based evaluation framework that adaptively probes the reasoning capabilities of mLLMs in deciphering meme harmfulness. Through multi-agent collaboration, AdamMeme provides comprehensive evaluations by iteratively updating the meme data with challenging samples, thereby exposing specific limitations in how mLLMs interpret harmfulness. Extensive experiments show that our framework systematically reveals the varying performance of different target mLLMs, offering in-depth, fine-grained analyses of model-specific weaknesses. Our code is available at https://github.com/Lbotirx/AdamMeme.

Image goal navigation requires two different skills: firstly, core navigation skills, including the detection of free space and obstacles, and taking decisions based on an internal representation; and secondly, computing directional information by comparing visual observations to the goal image. Current state-of-the-art methods either rely on dedicated image-matching, or pre-training of computer vision modules on relative pose estimation. In this paper, we study whether this task can be efficiently solved with end-to-end training of full agents with RL, as has been claimed by recent work. A positive answer would have impact beyond Embodied AI and allow training of relative pose estimation from reward for navigation alone. In this large experimental study we investigate the effect of architectural choices like late fusion, channel stacking, space-to-depth projections and cross-attention, and their role in the emergence of relative pose estimators from navigation training. W e show that the success of recent methods is influenced up to a certain extent by simulator settings, leading to shortcuts in simulation. However, we also show that these capabilities can be transferred to more realistic setting, up to some extent. W e also find evidence for correlations between navigation performance and probed (emerging) relative pose estimation performance, an important sub skill.

In this paper, we present a novel funnel-based tracking control algorithm for robotic systems with unknown dynamics and prescribed input constraints. The Euler-Lagrange formulation, a common modeling approach for robotic systems, has been adopted in this study to address the trade-off between performance and actuator safety. We establish feasibility conditions that ensure tracking errors evolve within predefined funnel bounds while maintaining bounded control efforts, a crucial consideration for robots with limited actuation capabilities. We propose two approximation-free control strategies for scenarios where these conditions are violated: one actively corrects the error, and the other stops further deviation. Finally, we demonstrate the robust performance and safety of the approach through simulations and experimental validations. This work represents a significant advancement in funnel-based control, enhancing its applicability to real-world robotics systems with input constraints.

Large language models (LLMs) have demonstrated impressive capabilities as autonomous agents with rapidly expanding applications in various domains. As these agents increasingly engage in socioeconomic interactions, identifying their potential for undesirable behavior becomes essential. In this work, we examine scenarios where they can choose to collude, defined as secretive cooperation that harms another party. To systematically study this, we investigate the behavior of LLM agents acting as sellers in simulated continuous double auction markets. Through a series of controlled experiments, we analyze how parameters such as the ability to communicate, choice of model, and presence of environmental pressures affect the stability and emergence of seller collusion. We find that direct seller communication increases collusive tendencies, the propensity to collude varies across models, and environmental pressures, such as oversight and urgency from authority figures, influence collusive behavior. Our findings highlight important economic and ethical considerations for the deployment of LLM-based market agents.

In this paper, we propose to incorporate the blackboard architecture into LLM multi-agent systems (MASs) so that (1) agents with various roles can share all the information and others' messages during the whole problem-solving process, (2) agents that will take actions are selected based on the current content of the blackboard, and (3) the selection and execution round is repeated until a consensus is reached on the blackboard. We develop the first implementation of this proposal and conduct experiments on commonsense knowledge, reasoning and mathematical datasets. The results show that our system can be competitive with the SOTA static and dynamic MASs by achieving the best average performance, and at the same time manage to spend less tokens. Our proposal has the potential to enable complex and dynamic problem-solving where well-defined structures or workflows are unavailable.

Multi-agent systems, which consist of multiple AI models interacting within a shared environment, are increasingly used for persona-based interactions. However, if not carefully designed, these systems can reinforce implicit biases in large language models (LLMs), raising concerns about fairness and equitable representation. We present MALIBU, a novel benchmark developed to assess the degree to which LLM-based multi-agent systems implicitly reinforce social biases and stereotypes. MALIBU evaluates bias in LLM-based multi-agent systems through scenario-based assessments. AI models complete tasks within predefined contexts, and their responses undergo evaluation by an LLM-based multi-agent judging system in two phases. In the first phase, judges score responses labeled with specific demographic personas (e.g., gender, race, religion) across four metrics. In the second phase, judges compare paired responses assigned to different personas, scoring them and selecting the superior response. Our study quantifies biases in LLM-generated outputs, revealing that bias mitigation may favor marginalized personas over true neutrality, emphasizing the need for nuanced detection, balanced fairness strategies, and transparent evaluation benchmarks in multi-agent systems.

Existing function-calling benchmarks focus on single-turn interactions. However, they overlook the complexity of real-world scenarios. To quantify how existing benchmarks address practical applications, we introduce DICE-SCORE, a metric that evaluates the dispersion of tool-related information such as function name and parameter values throughout the dialogue. Analyzing existing benchmarks through DICE-SCORE reveals notably low scores, highlighting the need for more realistic scenarios. To address this gap, we present DICE-BENCH, a framework that constructs practical function-calling datasets by synthesizing conversations through a tool graph that maintains dependencies across rounds and a multi-agent system with distinct personas to enhance dialogue naturalness. The final dataset comprises 1,607 high-DICE-SCORE instances. Our experiments on 19 LLMs with DICE-BENCH show that significant advances are still required before such models can be deployed effectively in real-world settings. Our code and data are all publicly available: https://snuhcc.github.io/DICE-Bench/.

Artificial intelligence holds great promise for expanding access to expert medical knowledge and reasoning. However, most evaluations of language models rely on static vignettes and multiple-choice questions that fail to reflect the complexity and nuance of evidence-based medicine in real-world settings. In clinical practice, physicians iteratively formulate and revise diagnostic hypotheses, adapting each subsequent question and test to what they've just learned, and weigh the evolving evidence before committing to a final diagnosis. To emulate this iterative process, we introduce the Sequential Diagnosis Benchmark, which transforms 304 diagnostically challenging New England Journal of Medicine clinicopathological conference (NEJM-CPC) cases into stepwise diagnostic encounters. A physician or AI begins with a short case abstract and must iteratively request additional details from a gatekeeper model that reveals findings only when explicitly queried. Performance is assessed not just by diagnostic accuracy but also by the cost of physician visits and tests performed. We also present the MAI Diagnostic Orchestrator (MAI-DxO), a model-agnostic orchestrator that simulates a panel of physicians, proposes likely differential diagnoses and strategically selects high-value, cost-effective tests. When paired with OpenAI's o3 model, MAI-DxO achieves 80% diagnostic accuracy--four times higher than the 20% average of generalist physicians. MAI-DxO also reduces diagnostic costs by 20% compared to physicians, and 70% compared to off-the-shelf o3. When configured for maximum accuracy, MAI-DxO achieves 85.5% accuracy. These performance gains with MAI-DxO generalize across models from the OpenAI, Gemini, Claude, Grok, DeepSeek, and Llama families. We highlight how AI systems, when guided to think iteratively and act judiciously, can advance diagnostic precision and cost-effectiveness in clinical care.

The remarkable advancements of vision and language foundation models in multimodal understanding, reasoning, and generation has sparked growing efforts to extend such intelligence to the physical world, fueling the flourishing of vision-language-action (VLA) models. Despite seemingly diverse approaches, we observe that current VLA models can be unified under a single framework: vision and language inputs are processed by a series of VLA modules, producing a chain of \textit{action tokens} that progressively encode more grounded and actionable information, ultimately generating executable actions. We further determine that the primary design choice distinguishing VLA models lies in how action tokens are formulated, which can be categorized into language description, code, affordance, trajectory, goal state, latent representation, raw action, and reasoning. However, there remains a lack of comprehensive understanding regarding action tokens, significantly impeding effective VLA development and obscuring future directions. Therefore, this survey aims to categorize and interpret existing VLA research through the lens of action tokenization, distill the strengths and limitations of each token type, and identify areas for improvement. Through this systematic review and analysis, we offer a synthesized outlook on the broader evolution of VLA models, highlight underexplored yet promising directions, and contribute guidance for future research, hoping to bring the field closer to general-purpose intelligence.

Patents contain rich technical knowledge that can inspire innovative product ideas, yet accessing and interpreting this information remains a challenge. This work explores the use of Large Language Models (LLMs) and autonomous agents to mine and generate product concepts from a given patent. In this work, we design Agent Ideate, a framework for automatically generating product-based business ideas from patents. We experimented with open-source LLMs and agent-based architectures across three domains: Computer Science, Natural Language Processing, and Material Chemistry. Evaluation results show that the agentic approach consistently outperformed standalone LLMs in terms of idea quality, relevance, and novelty. These findings suggest that combining LLMs with agentic workflows can significantly enhance the innovation pipeline by unlocking the untapped potential of business idea generation from patent data.