HKUST NLP
We track 4 benchmarks that HKUST NLP made or helped make.
3 in use 3 with public code 4 with a leaderboard here
Benchmarks
| Name | Kind | Released | Best score | Status |
|---|---|---|---|---|
| Toolathlon 108 long-horizon tasks across 32 apps through about 600 MCP tools. Execution-based checkers inspect final environment state. | Apps and MCP servers | Oct 2025 |  56.5% | In use |
| Toolathlon-Verified 108 long-horizon tasks across 32 apps through about 600 MCP tools. Execution-based checkers inspect final environment state. | Apps and MCP servers | Jun 2026 | 78.4% | In use |
| LOCA-Bench Agentic tasks whose environment state is scaled to grow context length while task semantics stay fixed. Success rate versus context length. | Reasoning over long documents | Feb 2026 | 21.3% | In use |
| AgentBoard LLM agents act in 9 partially observable, multi-turn environments that cover embodied tasks, games, web browsing and tool use. Besides final success, AgentBoard reports a progress rate, the share of annotated subgoals the agent reaches. An open toolkit breaks results down by skill. | General assistants | Jan 2024 | 48.9% | Not in cards |