Primary sourceshao__meng
View sourceEvent ·Single source
HarnessDev基准评测LLM创建Agent Harness能力
ByteDance Seed团队发布HarnessDev基准，评测LLM能否自行创建和演化Agent执行基础设施。该基准包含Creation和Evolution两个阶段，涉及Opus 4.8、GPT-5.5等6个模型在4领域5基准的表现。结果显示代码领域落后人类参考10.7分，状态记忆几乎全是死代码，且成本与效果不相关，同等分数token消耗可差7-19倍。
Current read
ByteDance团队推出HarnessDev基准，首次评测LLM能否自己创建Agent执行框架，发现模型间harness迁移性极差。
2 sources73° AI heatLast updated 2026/9/4 00:41:43
Evidence
2 sourcesRelated source 1elvis
View sourceConflict check
Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.