PRISM: Precision and contact-rich Real-world Industrial Skill dataset with Multimodal sensing
为解决工业装配中精准控制和多模态反馈不足的问题，通过收集包含多种操作任务的PRISM数据集，利用多视角RGB-D、力/扭矩、触觉等多模态传感信息。
0 citationsRead paper
Selected work
Representative Papers
OpenAgenet/OAN: Open Infrastructure for Trusted Agent Interconnection
This work addresses the lack of identity provenance, governance-state verification, authorization proofs, and pre-connection trust evidence among agents in open multi-operator networks by proposing the first unified trust architecture designed for open environments. The architecture establishes a protocol-agnostic, decentralized yet governance-controlled trust-layer infrastructure through mechanisms including root governance identity onboarding, registrar-assisted access, root validation package dissemination, authorization-aware discovery, and signature-based trusted invocation. This enables verifiable cross-domain interactions encompassing identity, governance, and authorization. A prototype implementation demonstrates the system’s performance and deployment feasibility, while clearly articulating a cooperative model and evolutionary pathway to provide a scalable trust foundation for open multi-agent networks.
0 citationsRead paper
OpenAgenet/OAN: Technical Architecture for Trust-Governed Agent Identity and Discovery
This work addresses the lack of a unified and trustworthy identity management and discovery mechanism for multi-agent systems in open environments by proposing a protocol-agnostic trust layer architecture. The architecture formalizes agent roles, registration, governance, and secure interactions through identity object modeling, a Root-governed lifecycle model, authorization-aware discovery mechanisms, and signed trusted invocations. Innovatively, it introduces Root verification bundles and protocol-independent security boundaries, enabling support for heterogeneous protocols such as MCP, A2A, and ANP. This approach facilitates verifiable identities, discoverability, and secure onboarding of agents across diverse frameworks, thereby establishing a general-purpose trust infrastructure for open multi-agent ecosystems.
0 citationsRead paper
GRAIL: A Deep-Granularity Hybrid Resonance Framework for Real-Time Agent Discovery via SLM-Enhanced Indexing
This work addresses the high discovery latency and low semantic precision in large-scale multi-agent collaboration caused by reliance on large language models (LLMs) or single-vector dense retrieval. To overcome these limitations, the authors propose GRAIL, a novel framework integrating three key innovations: fine-tuned small language models (SLMs) for millisecond-level capability label prediction, pseudo-document expansion to enrich semantic density, and a MaxSim-based fine-grained resonance matching mechanism. Evaluated on the AgentTaxo-9K dataset, GRAIL achieves an end-to-end discovery latency under 400 ms—79× faster than LLM baselines—while significantly outperforming conventional dense retrieval methods in Recall@10, thereby enabling highly accurate and real-time agent discovery.
0 citationsRead paper
Decoupling Speculation from Merit: The Identity-Bound Asset Integrity Model (IBAIM) for Sustainable Web3 Gaming
This study addresses the “death spiral” in Web3 gaming driven by speculative bubbles, capital monopolization, and inflation. It argues that a sustainable open-game economy must satisfy three core conditions: resistance to Sybil attacks, resilience against capital dominance, and immunity to inflationary saturation. To this end, the authors propose the Identity-Bound Asset Integrity Model (IBAIM), which uniquely integrates privacy-preserving biometric identity verification with an asymmetric utility decay mechanism. Leveraging zero-knowledge biometric hashing, account abstraction, and zk-PoI (zero-knowledge Proof of Identity), IBAIM anchors asset utility to verifiable human identities, thereby decoupling speculation from genuine gameplay achievements. Furthermore, a thermodynamic entropy–inspired asset degradation mechanism disentangles speculative behavior from endogenous value creation. Empirical results demonstrate that moderately constraining asset liquidity significantly enhances long-term system stability, offering both theoretical grounding and a practical pathway toward sustainable GameFi ecosystems.
0 citationsRead paper
Recent publications
Latest Papers
PRISM: Precision and contact-rich Real-world Industrial Skill dataset with Multimodal sensing
为解决工业装配中精准控制和多模态反馈不足的问题，通过收集包含多种操作任务的PRISM数据集，利用多视角RGB-D、力/扭矩、触觉等多模态传感信息。
0 citationsRead paper
OpenAgenet/OAN: Open Infrastructure for Trusted Agent Interconnection
This work addresses the lack of identity provenance, governance-state verification, authorization proofs, and pre-connection trust evidence among agents in open multi-operator networks by proposing the first unified trust architecture designed for open environments. The architecture establishes a protocol-agnostic, decentralized yet governance-controlled trust-layer infrastructure through mechanisms including root governance identity onboarding, registrar-assisted access, root validation package dissemination, authorization-aware discovery, and signature-based trusted invocation. This enables verifiable cross-domain interactions encompassing identity, governance, and authorization. A prototype implementation demonstrates the system’s performance and deployment feasibility, while clearly articulating a cooperative model and evolutionary pathway to provide a scalable trust foundation for open multi-agent networks.
0 citationsRead paper
OpenAgenet/OAN: Technical Architecture for Trust-Governed Agent Identity and Discovery
This work addresses the lack of a unified and trustworthy identity management and discovery mechanism for multi-agent systems in open environments by proposing a protocol-agnostic trust layer architecture. The architecture formalizes agent roles, registration, governance, and secure interactions through identity object modeling, a Root-governed lifecycle model, authorization-aware discovery mechanisms, and signed trusted invocations. Innovatively, it introduces Root verification bundles and protocol-independent security boundaries, enabling support for heterogeneous protocols such as MCP, A2A, and ANP. This approach facilitates verifiable identities, discoverability, and secure onboarding of agents across diverse frameworks, thereby establishing a general-purpose trust infrastructure for open multi-agent ecosystems.
0 citationsRead paper
GRAIL: A Deep-Granularity Hybrid Resonance Framework for Real-Time Agent Discovery via SLM-Enhanced Indexing
This work addresses the high discovery latency and low semantic precision in large-scale multi-agent collaboration caused by reliance on large language models (LLMs) or single-vector dense retrieval. To overcome these limitations, the authors propose GRAIL, a novel framework integrating three key innovations: fine-tuned small language models (SLMs) for millisecond-level capability label prediction, pseudo-document expansion to enrich semantic density, and a MaxSim-based fine-grained resonance matching mechanism. Evaluated on the AgentTaxo-9K dataset, GRAIL achieves an end-to-end discovery latency under 400 ms—79× faster than LLM baselines—while significantly outperforming conventional dense retrieval methods in Recall@10, thereby enabling highly accurate and real-time agent discovery.
0 citationsRead paper
Decoupling Speculation from Merit: The Identity-Bound Asset Integrity Model (IBAIM) for Sustainable Web3 Gaming
This study addresses the “death spiral” in Web3 gaming driven by speculative bubbles, capital monopolization, and inflation. It argues that a sustainable open-game economy must satisfy three core conditions: resistance to Sybil attacks, resilience against capital dominance, and immunity to inflationary saturation. To this end, the authors propose the Identity-Bound Asset Integrity Model (IBAIM), which uniquely integrates privacy-preserving biometric identity verification with an asymmetric utility decay mechanism. Leveraging zero-knowledge biometric hashing, account abstraction, and zk-PoI (zero-knowledge Proof of Identity), IBAIM anchors asset utility to verifiable human identities, thereby decoupling speculation from genuine gameplay achievements. Furthermore, a thermodynamic entropy–inspired asset degradation mechanism disentangles speculative behavior from endogenous value creation. Empirical results demonstrate that moderately constraining asset liquidity significantly enhances long-term system stability, offering both theoretical grounding and a practical pathway toward sustainable GameFi ecosystems.
0 citationsRead paper