arXiv subjects

Siyao Zhao

Publications and source records attributed to Siyao Zhao.

6 recordsLinked to original sources

arXiv2026

MARTI-MARS$^2$: Scaling Multi-Agent Self-Search via Reinforcement Learning for Code Generation

While the complex reasoning capability of Large Language Models (LLMs) has attracted significant attention, single-agent systems often encounter inherent performance ceilings in complex tasks such as code generation. Multi-agent collaboration offers a promising avenue to transcend these boundaries. However, existing frameworks typically rely on prompt-based test-time interactions or multi-role configurations trained with homogeneous parameters, limiting error correction capabilities and strategic diversity. In this paper, we propose a Multi-Agent Reinforced Training and Inference Framework with Self-Search Scaling (MARTI-MARS2), which integrates policy learning with multi-agent tree search by formulating the multi-agent collaborative exploration process as a dynamic and learnable environment. By allowing agents to iteratively explore and refine within the environment, the framework facilitates evolution from parameter-sharing homogeneous multi-role training to heterogeneous multi-agent training, breaking through single-agent capability limits. We also introduce an efficient inference strategy MARTI-MARS2-T+ to fully exploit the scaling potential of multi-agent collaboration at test time. We conduct extensive experiments across varied model scales (8B, 14B, and 32B) on challenging code generation benchmarks. Utilizing two collaborating 32B models, MARTI-MARS2 achieves 77.7%, outperforming strong baselines like GPT-5.1. Furthermore, MARTI-MARS2 reveals a novel scaling law: shifting from single-agent to homogeneous multi-role and ultimately to heterogeneous multi-agent paradigms progressively yields higher RL performance ceilings, robust TTS capabilities, and greater policy diversity, suggesting that policy diversity is critical for scaling intelligence via multi-agent reinforcement learning.

cs.LG↗

arXiv2025

TacEleven: generative tactic discovery for football open play

Creating offensive advantages during open play is fundamental to football success. However, due to the highly dynamic and long-sequence nature of open play, the potential tactic space grows exponentially as the sequence progresses, making automated tactic discovery extremely challenging. To address this, we propose TacEleven, a generative framework for football open-play tactic discovery developed in close collaboration with domain experts from AJ Auxerre, designed to assist coaches and analysts in tactical decision-making. TacEleven consists of two core components: a language-controlled tactical generator that produces diverse tactical proposals, and a multimodal large language model-based tactical critic that selects the optimal proposal aligned with a high-level stylistic tactical instruction. The two components enables rapid exploration of tactical proposals and discovery of alternative open-play offensive tactics. We evaluate TacEleven across three tasks with progressive tactical complexity: counterfactual exploration, single-step discovery, and multi-step discovery, through both quantitative metrics and a questionnaire-based qualitative assessment. The results show that the TacEleven-discovered tactics exhibit strong realism and tactical creativity, with 52.50% of the multi-step tactical alternatives rated adoptable in real-world elite football scenarios, highlighting the framework's ability to rapidly generate numerous high-quality tactics for complex long-sequence open-play situations. TacEleven demonstrates the potential of creatively leveraging domain data and generative models to advance tactical analysis in sports.

stat.AP↗

arXiv2025

Vision-Based Generic Potential Function for Policy Alignment in Multi-Agent Reinforcement Learning

Guiding the policy of multi-agent reinforcement learning to align with human common sense is a difficult problem, largely due to the complexity of modeling common sense as a reward, especially in complex and long-horizon multi-agent tasks. Recent works have shown the effectiveness of reward shaping, such as potential-based rewards, to enhance policy alignment. The existing works, however, primarily rely on experts to design rule-based rewards, which are often labor-intensive and lack a high-level semantic understanding of common sense. To solve this problem, we propose a hierarchical vision-based reward shaping method. At the bottom layer, a visual-language model (VLM) serves as a generic potential function, guiding the policy to align with human common sense through its intrinsic semantic understanding. To help the policy adapts to uncertainty and changes in long-horizon tasks, the top layer features an adaptive skill selection module based on a visual large language model (vLLM). The module uses instructions, video replays, and training records to dynamically select suitable potential function from a pre-designed pool. Besides, our method is theoretically proven to preserve the optimal policy. Extensive experiments conducted in the Google Research Football environment demonstrate that our method not only achieves a higher win rate but also effectively aligns the policy with human common sense.

cs.AI↗

arXiv2023

Implementation and performances of the IPbus protocol for the JUNO Large-PMT readout electronics

The Jiangmen Underground Neutrino Observatory (JUNO) is a large neutrino detector currently under construction in China. Thanks to the tight requirements on its optical and radio-purity properties, it will be able to perform leading measurements detecting terrestrial and astrophysical neutrinos in a wide energy range from tens of keV to hundreds of MeV. A key requirement for the success of the experiment is an unprecedented 3% energy resolution, guaranteed by its large active mass (20 kton) and the use of more than 20,000 20-inch photo-multiplier tubes (PMTs) acquired by high-speed, high-resolution sampling electronics located very close to the PMTs. As the Front-End and Read-Out electronics is expected to continuously run underwater for 30 years, a reliable readout acquisition system capable of handling the timestamped data stream coming from the Large-PMTs and permitting to simultaneously monitor and operate remotely the inaccessible electronics had to be developed. In this contribution, the firmware and hardware implementation of the IPbus based readout protocol will be presented, together with the performances measured on final modules during the mass production of the electronics.

physics.ins-det↗

arXiv2023

Mass testing of the JUNO experiment 20-inch PMTs readout electronics

The Jiangmen Underground Neutrino Observatory (JUNO) is a multi-purpose, large size, liquid scintillator experiment under construction in China. JUNO will perform leading measurements detecting neutrinos from different sources (reactor, terrestrial and astrophysical neutrinos) covering a wide energy range (from 200 keV to several GeV). This paper focuses on the design and development of a test protocol for the 20-inch PMT underwater readout electronics, performed in parallel to the mass production line. In a time period of about ten months, a total number of 6950 electronic boards were tested with an acceptance yield of 99.1%.

physics.ins-det↗

arXiv2022

Validation and integration tests of the JUNO 20-inch PMTs readout electronics

The Jiangmen Underground Neutrino Observatory (JUNO) is a large neutrino detector currently under construction in China. JUNO will be able to study the neutrino mass ordering and to perform leading measurements detecting terrestrial and astrophysical neutrinos in a wide energy range, spanning from 200 keV to several GeV. Given the ambitious physics goals of JUNO, the electronic system has to meet specific tight requirements, and a thorough characterization is required. The present paper describes the tests performed on the readout modules to measure their performances.

physics.ins-det↗