Phala Research

我们在可信 AI 方向的研究——可信执行环境、安全 GPU 推理，以及您可自行验证的系统。

最新

First page of The Serialized Bridge: Understanding and Recovering LLM Serving Performance under Blackwell GPU Confidential Computing

GPU Confidential Computing · Jun 2026

The Serialized Bridge: Understanding and Recovering LLM Serving Performance under Blackwell GPU Confidential Computing

Locates the CVM–GPU bridge — not GPU compute — as the real cost of confidential LLM serving, and recovers up to 92% of lost throughput.

Hang Yin, Kevin Wang阅读论文

全部论文

First page of Implementing Kubernetes Pod-Level Remote Attestation for Confidential Workloads on dstack

TEE Infrastructure

Implementing Kubernetes Pod-Level Remote Attestation for Confidential Workloads on dstack

Yang Yang, Kevin Wang, +5Jun 2026

First page of LegalCiteBench: Evaluating Citation Reliability in Legal Language Models

AI Safety & Evaluation

LegalCiteBench: Evaluating Citation Reliability in Legal Language Models

Sijia Chen, Hang Yin, Shunfan ZhouMay 2026

First page of PASS: A Provenanced Access Subaccount System for Blockchain Wallets

Verifiable Systems

PASS: A Provenanced Access Subaccount System for Blockchain Wallets

Jay Yu, Shunfan Zhou, +2Apr 2026

First page of Persistent BitTorrent Trackers

Verifiable Systems

Persistent BitTorrent Trackers

François-Xavier Wicht, Zhengwei Tong, +3Nov 2025

First page of Dstack: A Zero Trust Framework for Confidential Containers

TEE Infrastructure

Dstack: A Zero Trust Framework for Confidential Containers

Shunfan Zhou, Kevin Wang, Hang YinSep 2025

First page of Confidential Computing on NVIDIA Hopper GPUs: A Performance Benchmark Study

GPU Confidential Computing

Confidential Computing on NVIDIA Hopper GPUs: A Performance Benchmark Study

Jianwei Zhu, Hang Yin, +3Sep 2024

隐私执行。可验证结果。

新闻通讯

© 2026 Hashforest Technology。保留所有权利。隐私 • 条款

Phala 研究 — 可信 AI、TEE 与安全 GPU 推理 | Phala