Overthinking Loops in Agents: A Structural Risk via MCP Tools

Tool-using LLM agents increasingly coordinate real workloads by selecting and chaining third-party tools based on text-visible metadata such as tool names, descriptions, and return messages. We show that this convenience creates a supply-chain attack surface: a malicious MCP tool server can be co-registered alongside normal tools and induce overthinking loops, where individually trivial or plausible tool calls compose into cyclic trajectories that inflate end-to-end tokens and latency without any single step looking abnormal. We formalize this as a structural overthinking attack, distinguishable from token-level verbosity, and implement 14 malicious tools across three servers that trigger repetition, forced refinement, and distraction. Across heterogeneous registries and multiple tool-capable models, the attack causes severe resource amplification (up to $142.4\times$ tokens) and can degrade task outcomes. Finally, we find that decoding-time concision controls do not reliably prevent loop induction, suggesting defenses should reason about tool-call structure rather than tokens alone.

MIMIC-D: Multi-modal Imitation for MultI-agent Coordination with Decentralized Diffusion Policies

As robots become more integrated in society, their ability to coordinate with other robots and humans on multi-modal tasks (those with multiple valid solutions) is crucial. We propose to learn such behaviors from expert demonstrations via imitation learning (IL). However, when expert demonstrations are multi-modal, standard IL approaches can struggle to capture the diverse strategies, hindering effective coordination. Diffusion models are known to be effective at handling complex multi-modal trajectory distributions in single-agent systems. Diffusion models have also excelled in multi-agent scenarios where multi-modality is more common and crucial to learning coordinated behaviors. Typically, diffusion-based approaches require a centralized planner or explicit communication among agents, but this assumption can fail in real-world scenarios where robots must operate independently or with agents like humans that they cannot directly communicate with. Therefore, we propose MIMIC-D, a Centralized Training, Decentralized Execution (CTDE) paradigm for multi-modal multi-agent imitation learning using diffusion policies. Agents are trained jointly with full information, but execute policies using only local information to achieve implicit coordination. We demonstrate in both simulation and hardware experiments that our method recovers multi-modal coordination behavior among agents in a variety of tasks and environments, while improving upon state-of-the-art baselines.

CRAFT: Coaching Reinforcement Learning Autonomously using Foundation Models for Multi-Robot Coordination Tasks

Multi-Agent Reinforcement Learning (MARL) provides a powerful framework for learning coordination in multi-agent systems. However, applying MARL to robotics still remains challenging due to high-dimensional continuous joint action spaces, complex reward design, and non-stationary transitions inherent to decentralized settings. On the other hand, humans learn complex coordination through staged curricula, where long-horizon behaviors are progressively built upon simpler skills. Motivated by this, we propose CRAFT: Coaching Reinforcement learning Autonomously using Foundation models for multi-robot coordination Tasks, a framework that leverages the reasoning capabilities of foundation models to act as a "coach" for multi-robot coordination. CRAFT automatically decomposes long-horizon coordination tasks into sequences of subtasks using the planning capability of Large Language Models (LLMs). In what follows, CRAFT trains each subtask using reward functions generated by LLM, and refines them through a Vision Language Model (VLM)-guided reward-refinement loop. We evaluate CRAFT on multi-quadruped navigation and bimanual manipulation tasks, demonstrating its capability to learn complex coordination behaviors. In addition, we validate the multi-quadruped navigation policy in real hardware experiments.