Seal-0

Emerging

2papers using it

2026first seen

The 'SEAL-0' dataset/benchmark is used to evaluate the performance of agentic frameworks on long-horizon tasks by providing a structured set of challenges that require complex reasoning and orchestration.

🔎 Find this dataset

Papers using Seal-0 (2)

Scaling the Horizon, Not the Parameters: Reaching Trillion-Parameter Performance with a 35B Agent2026

ROMA: Recursive Open Meta-Agent Framework for Long-Horizon Multi-Agent Systems2026