工作区 D:\Research\DoneBench。你不是独自在代码库中工作，不要回退他人改动。任务：Reviewer 2 / 顶会相关工作差距分析。请联…
- Stable ID:
codex-user-prompt:e61e0cf823d613e0
- Source kind:
codex-session-user
- Category:
research-workflow
- Timestamp:
2026-05-07T14:41:39.809Z
- Semantic hash:
e61e0cf823d613e0ddccc22fda5db3b0c479cce16535ba41a529baed68f34b9a
- Public handling: selected full-text prompt with secret filtering.
Prompt Text
工作区 D:\Research\DoneBench。你不是独自在代码库中工作，不要回退他人改动。任务：Reviewer 2 / 顶会相关工作差距分析。请联网查 WebArena、OSWorld、WorkArena/WorkArena++、tau-bench、AgentBoard、AgentEval/DeepEval、CodeSpecBench/RESTestBench、VerifyLLM 等，聚焦 benchmark artifact 模块、规模、创新点边界。职责范围：reports/reviewer2.md、reports/related_work_gap_analysis.md、paper/sections/related_work.tex 可编辑。目标：指出我们还缺什么、哪些创新点必须强调、哪些地方容易被说成 toy/拼接。最后报告引用来源和改动文件。
Reuse Notes
- EXTRACTED: This is a selected Codex prompt or automation prompt from the local Codex corpus.
- INFERRED: Future agents can reuse its structure, constraints, and acceptance criteria when creating similar Codex workflows.