Recent advances in text-guided diffusion models have revolutionized conditional image generation, yet they struggle to synthesize complex scenes with multiple objects due to imprecise spatial grounding and limited scalability. We address these challenges through two key modules: 1) Janus-Pro-driven Prompt Parsing, a prompt-layout parsing module that bridges text understanding and layout generation via a compact 1B-parameter architecture, and 2) MIGLoRA, a parameter-efficient plug-in integrating Low-Rank Adaptation (LoRA) into UNet (SD1.5) and DiT (SD3) backbones. MIGLoRA is capable of preserving the base model’s parameters and ensuring plug-and-play adaptability, minimizing architectural intrusion while enabling efficient fine-tuning. To support a comprehensive evaluation, we create DescripBox and DescripBox-1024, benchmarks that span diverse scenes and resolutions. The proposed method achieves state-of-the-art performance on COCO and LVIS benchmarks while maintaining parameter efficiency, demonstrating superior layout fidelity and scalability for open-world synthesis. Here is our project page: https://github.com/FanQi-AI/MIGLoRA.

The rapid advancement of multi-agent reinforcement learning (MARL) has given rise to diverse training paradigms to learn the policies of each agent in the multi-agent system. The paradigms of decentralized training and execution (DTDE) and centralized training with decentralized execution (CTDE) have been proposed and widely applied. However, as the number of agents increases, the inherent limitations of these frameworks significantly degrade the performance metrics, such as win rate, total reward, etc. To reduce the influence of the increasing number of agents on the performance metrics, we propose a novel training paradigm of grouped training decentralized execution (GTDE). This framework eliminates the need for a centralized module and relies solely on local information, effectively meeting the training requirements of large-scale multi-agent systems. Specifically, we first introduce an adaptive grouping module, which divides each agent into different groups based on their observation history. To implement end-to-end training, GTDE uses Gumbel-Sigmoid for efficient point-to-point sampling on the grouping distribution while ensuring gradient backpropagation. To adapt to the uncertainty in the number of members in a group, two methods are used to implement a group information aggregation module that merges member information within the group. Empirical results show that in a cooperative environment with 495 agents, GTDE increased the total reward by an average of 382% compared to the baseline. In a competitive environment with 64 agents, GTDE achieved a 100% win rate against the baseline.

While Large Language Models (LLMs) show promise for task planning, their efficacy diminishes in complex, long-horizon tasks within dynamic, partially observable environments, primarily due to challenges in long-term reasoning and effective adaptation from experience. A key limitation of current approaches is the insufficient utilization of historical trajectory information. To overcome these challenges in the context of Partially Observable Markov Decision Process (POMDP) planning, this paper introduces DMA-MCTS (Dynamic Memory-Augmented Monte-Carlo Tree Search), a framework that integrates Monte Carlo Tree Search (MCTS) with LLMs, augmented by a novel dynamic memory and reflection system. The core technical contributions include: (1) a dual-layer semantic memory repository enabling efficient context-aware retrieval of past experiences; (2) a memory-enhanced UCT selection strategy biased by historical Q-values to guide search; and (3) a differentiated reflection mechanism employing LLMs to extract generalizable knowledge from both successful and failed trajectories. Comprehensive evaluations conducted on complex object rearrangement tasks within the VirtualHome simulator demonstrate that DMA-MCTS significantly outperforms relevant baselines, including standard LLM-MCTS approaches, in terms of task success rate, generalization capabilities, and planning efficiency. These results underscore the critical importance of integrating structured dynamic memory and systematic reflection mechanisms for developing highly adaptive and effective LLM-based agents capable of tackling long-horizon planning problems.