Efficiency is increasingly important for Large Language Model (LLM)-based multi-agent systems (MAS), as larger models and more agents introduce substantial execution costs. Recent methods aim to make MAS cheaper by pruning agents, removing communication edges, or searching for compact structures. However, we argue that existing evaluations may overestimate their true ability to improve MAS efficiency. Reported gains are often measured under method-specific prompts and starting topologies, making them difficult to attribute to the proposed structural changes. Moreover, many reported successes appear in non-MAS-demanding settings, where a single agent or a randomly pruned system can already preserve strong performance. To study these issues, we introduce a controlled and MAS-demanding diagnostic benchmark for representative MAS efficiency methods. We evaluate methods under a shared backbone model, agent registry, and runtime, across controlled variations in topology, scale, depth, and tool use. Our analysis shows that many reported gains are setup-dependent and may arise from structural collapse, disabled tool pathways, or starting systems where random pruning already preserves accuracy, rather than robust improvements in MAS efficiency.

Text style transfer (TST) modifies a source sentence to match a target style while preserving its semantics. While existing models perform well on simple styles like sentiment and formality, they struggle with complex, entangled styles such as poetry and brand-specific tones, which require advanced operations to disentangle content and style. We propose a multi-agent self-check framework that contains a large language model (LLM) as a planner for disentangling subtasks and expert agents for executing the subtasks. This training-free multi-agent framework decomposes TST into manageable components, enabling iterative refinement through a self-check module that balances style adherence and content preservation. Experiments on both simple and complex style datasets show our framework significantly improves style strength and content preservation, with strong adaptability in few-shot settings.

Retrieval-augmented generation enhances large language models (LLMs) by incorporating relevant information from external knowledge sources. This enables LLMs to adapt to specific domains and mitigate hallucinations in knowledge-intensive tasks. However, existing retrievers are often misaligned with LLMs due to separate training processes and the inherent black-box nature of LLMs. To address this challenge, we propose ARL2, a retriever learning technique that harnesses LLMs as labelers. ARL2 leverages LLMs to annotate and score adaptive relevance evidence, enabling the retriever to learn from robust LLM supervision. Furthermore, ARL2 incorporates a self-training strategy to minimize the cost of API calls. Extensive experiments demonstrate the effectiveness of ARL2, achieving accuracy improvements of 5.4% on NQ and 4.6% on MMLU compared to the state-of-the-art methods. Additionally, ARL2 exhibits robust transfer learning capabilities and strong zero-shot generalization abilities.

The generalization problem on KBQA has drawn considerable attention. Existing research suffers from the generalization issue brought by the entanglement in the coarse-grained modeling of the logical expression, or inexecutability issues due to the fine-grained modeling of disconnected classes and relations in real KBs. We propose a Fine-to-Coarse Composition framework for KBQA (FC-KBQA) to both ensure the generalization ability and executability of the logical expression. The main idea of FC-KBQA is to extract relevant fine-grained knowledge components from KB and reformulate them into middle-grained knowledge pairs for generating the final logical expressions. FC-KBQA derives new state-of-the-art performance on GrailQA and WebQSP, and runs 4 times faster than the baseline. Our code is now available at GitHub https://github.com/RUCKBReasoning/FC-KBQA.