arXiv subjects

Shijian Wang

Publications and source records attributed to Shijian Wang.

26 records · Page 2Linked to original sources

arXiv2025

Agent2World: Learning to Generate Symbolic World Models via Adaptive Multi-Agent Feedback

Symbolic world models (e.g., PDDL domains or executable simulators) are central to model-based planning, but training LLMs to generate such world models is limited by the lack of large-scale verifiable supervision. Current approaches rely primarily on static validation methods that fail to catch behavior-level errors arising from interactive execution. In this paper, we propose Agent2World, a tool-augmented multi-agent framework that achieves strong inference-time world-model generation and also serves as a data engine for supervised fine-tuning, by grounding generation in multi-agent feedback. Agent2World follows a three-stage pipeline: (i) A Deep Researcher agent performs knowledge synthesis by web searching to address specification gaps; (ii) A Model Developer agent implements executable world models; And (iii) a specialized Testing Team conducts adaptive unit testing and simulation-based validation. Agent2World demonstrates superior inference-time performance across three benchmarks spanning both Planning Domain Definition Language (PDDL) and executable code representations, achieving consistent state-of-the-art results. Beyond inference, Testing Team serves as an interactive environment for the Model Developer, providing behavior-aware adaptive feedback that yields multi-turn training trajectories. The model fine-tuned on these trajectories substantially improves world-model generation, yielding an average relative gain of 30.95% over the same model before training. Project page: https://agent2world.github.io.

cs.AI↗

arXiv2025

Synthetic Curriculum Reinforces Compositional Text-to-Image Generation

Text-to-Image (T2I) generation has long been an open problem, with compositional synthesis remaining particularly challenging. This task requires accurate rendering of complex scenes containing multiple objects that exhibit diverse attributes as well as intricate spatial and semantic relationships, demanding both precise object placement and coherent inter-object interactions. In this paper, we propose a novel compositional curriculum reinforcement learning framework named CompGen that addresses compositional weakness in existing T2I models. Specifically, we leverage scene graphs to establish a novel difficulty criterion for compositional ability and develop a corresponding adaptive Markov Chain Monte Carlo graph sampling algorithm. This difficulty-aware approach enables the synthesis of training curriculum data that progressively optimize T2I models through reinforcement learning. We integrate our curriculum learning approach into Group Relative Policy Optimization (GRPO) and investigate different curriculum scheduling strategies. Our experiments reveal that CompGen exhibits distinct scaling curves under different curriculum scheduling strategies, with easy-to-hard and Gaussian sampling strategies yielding superior scaling performance compared to random sampling. Extensive experiments demonstrate that CompGen significantly enhances compositional generation capabilities for both diffusion-based and auto-regressive T2I models, highlighting its effectiveness in improving the compositional T2I generation systems.

cs.CV↗

arXiv2025

Video-Thinker: Sparking "Thinking with Videos" via Reinforcement Learning

Recent advances in image reasoning methods, particularly "Thinking with Images", have demonstrated remarkable success in Multimodal Large Language Models (MLLMs); however, this dynamic reasoning paradigm has not yet been extended to video reasoning tasks. In this paper, we propose Video-Thinker, which empowers MLLMs to think with videos by autonomously leveraging their intrinsic "grounding" and "captioning" capabilities to generate reasoning clues throughout the inference process. To spark this capability, we construct Video-Thinker-10K, a curated dataset featuring autonomous tool usage within chain-of-thought reasoning sequences. Our training strategy begins with Supervised Fine-Tuning (SFT) to learn the reasoning format, followed by Group Relative Policy Optimization (GRPO) to strengthen this reasoning capability. Through this approach, Video-Thinker enables MLLMs to autonomously navigate grounding and captioning tasks for video reasoning, eliminating the need for constructing and calling external tools. Extensive experiments demonstrate that Video-Thinker achieves significant performance gains on both in-domain tasks and challenging out-of-domain video reasoning benchmarks, including Video-Holmes, CG-Bench-Reasoning, and VRBench. Our Video-Thinker-7B substantially outperforms existing baselines such as Video-R1 and establishes state-of-the-art performance among 7B-sized MLLMs.

cs.CV↗

arXiv2025

Investigating the Scaling Effect of Instruction Templates for Training Multimodal Language Model

Current multimodal language model (MLM) training approaches overlook the influence of instruction templates. Previous research deals with this problem by leveraging hand-crafted or model-generated templates, failing to investigate the scaling effect of instruction templates on MLM training. In this work, we propose a programmatic instruction template generator capable of producing over 15K unique instruction templates by filling randomly sampled positional synonyms into weighted sampled meta templates, enabling us to comprehensively explore MLM's performance across various template scales in the training process. Our investigation into scaling instruction templates for MLM training demonstrates that MLM capabilities do not consistently improve with increasing template scale. Instead, optimal performance is achieved at a medium template scale. Models trained with data augmented at the optimal template scale achieve performance gains of up to 10% over those trained on the original data and achieve the best overall performance compared with the similar-scale MLMs tuned on at most 75 times the scale of our augmented dataset. The code will be publicly available at https://github.com/shijian2001/TemplateScaling.

cs.CV↗

arXiv2025

Attributed Synthetic Data Generation for Zero-shot Domain-specific Image Classification

Zero-shot domain-specific image classification is challenging in classifying real images without ground-truth in-domain training examples. Recent research involved knowledge from texts with a text-to-image model to generate in-domain training images in zero-shot scenarios. However, existing methods heavily rely on simple prompt strategies, limiting the diversity of synthetic training images, thus leading to inferior performance compared to real images. In this paper, we propose AttrSyn, which leverages large language models to generate attributed prompts. These prompts allow for the generation of more diverse attributed synthetic images. Experiments for zero-shot domain-specific image classification on two fine-grained datasets show that training with synthetic images generated by AttrSyn significantly outperforms CLIP's zero-shot classification under most situations and consistently surpasses simple prompt strategies.

cs.CV↗

arXiv2025

Adaptive In-conversation Team Building for Language Model Agents

Leveraging multiple large language model (LLM) agents has shown to be a promising approach for tackling complex tasks, while the effective design of multiple agents for a particular application remains an art. It is thus intriguing to answer a critical question: Given a task, how can we build a team of LLM agents to solve it effectively? Our new adaptive team-building paradigm offers a flexible solution, realized through a novel agent design named Captain Agent. It dynamically forms and manages teams for each step of a task-solving process, utilizing nested group conversations and reflection to ensure diverse expertise and prevent stereotypical outputs, allowing for a flexible yet structured approach to problem-solving. A comprehensive evaluation across six real-world scenarios demonstrates that Captain Agent significantly outperforms existing multi-agent methods with 21.94% improvement in average accuracy, providing outstanding performance without requiring task-specific prompt engineering. Our exploration of different backbone LLM and cost analysis further shows that Captain Agent can improve the conversation quality of weak LLM and achieve competitive performance with extremely low cost, which illuminates the application of multi-agent systems.

cs.CL↗

arXiv2024

In situ Qubit Frequency Tuning Circuit for Scalable Superconducting Quantum Computing: Scheme and Experiment

Frequency tunable qubit plays a significant role for scalable superconducting quantum processors. The state-of-the-art room-temperature electronics for tuning qubit frequency suffers from unscalable limit, such as heating problem, linear growth of control cables, etc. Here we propose a scalable scheme to tune the qubit frequency by using in situ superconducting circuit, which is based on radio frequency superconducting quantum interference device (rf-SQUID). We demonstrate both theoretically and experimentally that the qubit frequency could be modulated by inputting several single pulses into rf-SQUID. Compared with the traditional scheme, our scheme not only solves the heating problem, but also provides the potential to exponentially reduce the number of cables inside the dilute refrigerator and the room-temperature electronics resource for tuning qubit frequency, which is achieved by a time-division-multiplex (TDM) scheme combining rf-SQUID with switch arrays. With such TDM scheme, the number of cables could be reduced from the usual $\sim 3n$ to $\sim \log_2{(3n)} + 1$ for two-dimensional quantum processors comprising $n$ qubits and $\sim 2n$ couplers. Our work paves the way for large-scale control of superconducting quantum processor.

quant-ph↗

arXiv2024

Demonstration of 3 V Programmable Josephson Junction Arrays Using Non-Integer-Multiple Logic

This article demonstrates a new kind of programmable logic for the representation of an integer that can be used for the programmable Josephson voltage standard. It can enable the numbers of junctions in most bits to be variable integer values, which is different from normal binary logic or ternary logic. Consequently, missing junctions due to superconducting short circuits can be tolerated under this logic. This logic can also have nearly the same segmentation efficiency as ternary logic. The completeness of the sequences using this logic is proven by the recursive method in mathematics in this paper. After that, a new algorithm for the representation of integers is presented according to the proven process, and an analysis of the number of fault-tolerant junctions for each bit is provided. Although the first and second bits are not tolerant to missing junctions, bits beyond these can tolerate one to hundreds of missing junctions. Due to the non-fixed multiples between the bits of the sequence, this logic is called non-integer-multiple logic. Finally, the design and fabrication of a 3 V programmable Josephson junction array using this logic are described, and the measurements and analysis of the characteristic parameters are presented.

cs.ET↗