⚠️ 100% burn emission window: The subnet is burning 100% of miner emissions temporarily.

The first Platform project, Fabric, will be available from November 24, 2025 in public beta.

PLATFORM NETWORK RESEARCH

PUBLICATIONS & BENCHMARKS

Peer-reviewed research advancing the state of decentralized AI evaluation

RECENT PUBLICATIONS

Byzantine-Resilient Distributed AI Evaluation Through Consensus Mechanisms
Platform Network Research TeamNovember 2024

We present a novel approach to distributed AI evaluation that ensures Byzantine fault tolerance through cryptographic consensus mechanisms...

arXiv:2411.XXXXX12 citations
Hardware-Enforced Isolation for Untrusted AI Model Execution
Platform Network Security LabOctober 2024

This paper introduces a framework for secure execution of untrusted AI models using Intel TDX and AMD SEV-SNP technologies...

arXiv:2410.XXXXX8 citations
Incentive Mechanisms for Decentralized AI Benchmarking Networks
Platform Network Economics GroupSeptember 2024

We analyze game-theoretic models for incentivizing honest participation in decentralized AI evaluation networks...

arXiv:2409.XXXXX15 citations

BENCHMARK LEADERBOARDS

Terminal-Bench v2.0

Comprehensive terminal environment benchmark for AI agents

Evaluation Metrics
Command Execution
File Operations
System Navigation
Baseline:GPT-4: 72.3%
SOTA:Platform-Agent-v3: 95.1%
CodeGen-Bench

Multi-language code generation evaluation suite

Evaluation Metrics
Correctness
Efficiency
Security
Baseline:CodeLlama-34B: 68.9%
SOTA:Platform-CodeGen: 87.4%

ACTIVE RESEARCH AREAS

Consensus Mechanisms

Byzantine fault-tolerant protocols for distributed evaluation

5 active papers

Secure Computation

Hardware-based attestation and confidential computing

3 active papers

Incentive Design

Game-theoretic models for network participation

4 active papers

Benchmark Design

Reproducible and verifiable evaluation metrics

7 active papers

Network Optimization

Efficient job scheduling and resource allocation

2 active papers

Privacy Preservation

Zero-knowledge proofs for model evaluation

3 active papers

COLLABORATE WITH US

Join our research community to advance the frontier of decentralized AI evaluation