Project report / GitHub evidence
OpenClaw ClawBench Model Card
这是可索引项目报告证据页：它保留 OpenClaw ClawBench Model Card 的源材料入口、机制线索和限制提醒；正文仍需 reader/editor 与 academic public-copy review 后才能当作最终结论引用。
OpenClaw ClawBench Model Card
| Field | Value |
|---|---|
| Repository | openclaw/clawbench |
| Category | Agent Harness 评测诊断 |
| Stars / forks snapshot | 97 / 18 |
| Language | Python |
| License | MIT |
| Raw capture | raw-github/openclaw_clawbench.md |
| Updated by | hourly public metadata update, 2026-05-24 |
1. Role in Self Evolve
把模型能力、插件栈、harness 配置和执行轨迹一起纳入评价，避免只看最终输出。
2. Working Principle
trace scoring + reliability metrics + seed-noise/capability-signal decomposition + dynamical-systems failure regime diagnostics。
3. Evidence Path
web GitHub page captured Core v1 2026-04-20 release notes and repository surface metadata.
4. Teaching Use
Use this card to explain whether a repository is a runtime, benchmark, harness-evolution loop, memory/skill substrate, or resource index. The key reading path is: raw capture -> classification row -> public site card -> project report -> aggregate GitHub analysis.
5. Limits
当前未运行 1,080-run sweep，也没有本地 GitNexus symbol graph；仅作为公开 metadata 与方法信号。