Harness Evals Model Card
这是可索引项目报告证据页：它保留 Harness Evals 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。
Harness Evals Model Card
| Field | Value |
|---|---|
| Repository | harness/harness-evals |
| Category | Agent Reliability Evaluation Framework |
| Stars / forks snapshot | 3 / 3 |
| Language | Python |
| License | Apache-2.0 |
| Raw capture | raw-github/harness_harness-evals.md |
| Updated by | hourly public metadata update, 2026-05-27 16:00 +0800 |
1. Role in Self Evolve
Harness Evals is an open-source evaluation framework for LLM agents, prompts, and structured outputs with normalized score semantics. It matters because self-evolving agents need repeatable harness control, measurable feedback loops, and reusable skill procedures before claiming stable improvement.
2. Working Principle
evaluate cases with normalized 0-1 scores -> configurable pass thresholds -> optional llm judged metrics and telemetry sinks -> regression export to CI observability pipelines
3. Evidence Path
web-observed GitHub page showed 3 stars, 3 forks, 86 commits, Apache-2.0 license, and README claims for normalized scoring, threshold-based pass/fail, optional LLM metrics, and OTLP/Langfuse integration. Shell GitHub API access remained blocked by DNS and local gh auth was invalid, so this card treats the snapshot as web-observed rather than API-verified.
4. Teaching Use
Use this card to explain Agent Reliability Evaluation Framework: it shows how harness/runtime/benchmark layers convert agent behavior into reproducible and auditable engineering workflows.
5. Limits
The repository was not cloned in this iteration; no benchmark run, plugin install, workflow execution, or agent loop experiment was executed. Counts and claims are visible public-page/search signals unless independently revalidated later.