Collect first agent research and market sources
This commit is contained in:
wuyang
@@ -0,0 +1,73 @@
|
||||
# Collection Run: 2026-07-08 First Sweep
|
||||
|
||||
status: completed
|
||||
|
||||
## Scope
|
||||
|
||||
- date: 2026-07-08
|
||||
- jobs: rolling 6-month window, Beijing-first Agent / LLM roles
|
||||
- papers: provisional source sweep from arXiv and official research/blog sources
|
||||
- industry: large-company Agent work, technical reports, engineering blogs, benchmarks
|
||||
|
||||
## Queries Used
|
||||
|
||||
Jobs:
|
||||
|
||||
```text
|
||||
site:joinbytedance.com 北京 Agent 大模型 LLM 招聘
|
||||
site:careers.tencent.com 北京 Agent 大模型 LLM 招聘
|
||||
site:talent.baidu.com 北京 Agent 大模型 LLM 招聘
|
||||
site:careers.alibaba.com 北京 Agent 大模型 LLM 招聘
|
||||
DeepSeek 招聘 Agent 北京 大模型 岗位
|
||||
site:jobs.bytedance.com 北京 智能体 大模型 Agent
|
||||
site:careers.tencent.com/jobdesc.html 北京 LLM Agent 桌面 智能伙伴
|
||||
```
|
||||
|
||||
Papers:
|
||||
|
||||
```text
|
||||
2026 LLM agent evaluation benchmark arxiv tool use memory
|
||||
2026 AI agent benchmark evaluation arxiv long horizon tasks
|
||||
2026 coding agent benchmark SWE-bench arxiv agent evaluation
|
||||
2026 multi-agent AI safety benchmark arxiv
|
||||
```
|
||||
|
||||
Industry:
|
||||
|
||||
```text
|
||||
site:openai.com/index agents 2026 technical report
|
||||
site:anthropic.com/news agent computer use 2026
|
||||
site:deepmind.google/blog agent 2026
|
||||
site:microsoft.com/en-us/research/blog agent 2026
|
||||
site:qwen.ai/blog agent Qwen 2026
|
||||
site:seed.bytedance.com/en agent 2026 ByteDance Seed
|
||||
site:github.com/deepseek-ai agent DeepSeek 2026
|
||||
```
|
||||
|
||||
## First-pass Method Notes
|
||||
|
||||
- Official pages are preferred for item files.
|
||||
- Search snippets are allowed only as weak evidence when official pages are JavaScript-rendered and not easily extractable.
|
||||
- For JD, source quality and location must be explicit; non-Beijing roles can be kept only when they reveal strong Agent demand signals.
|
||||
- For papers, items are marked `queued` or `skimmed`; source approval remains separate from item relevance.
|
||||
- For industry material, `source_quality` and `evidence_level` are deliberately separate.
|
||||
|
||||
## Output
|
||||
|
||||
- jobs: 6 starter items
|
||||
- papers: 6 starter items
|
||||
- industry: 9 starter items
|
||||
- scripts: item creation, index generation, URL checking
|
||||
|
||||
## Summary
|
||||
|
||||
- total items: 21
|
||||
- by collection: jobs 6, papers 6, industry 9
|
||||
- strongest recurring topics: agent evaluation, memory, coding agents, safety, enterprise deployment
|
||||
- first tool output: `data/index.json` and `data/summary.json`
|
||||
|
||||
## Follow-up
|
||||
|
||||
- Deepen official JD extraction for DeepSeek, ByteDance and Tencent.
|
||||
- Confirm approved paper sources before doing a larger paper sweep.
|
||||
- Add Anthropic and more Qwen/OpenAI/DeepSeek industry items in the next run.
|
||||
+1016
File diff suppressed because it is too large
Load Diff
@@ -0,0 +1,224 @@
|
||||
{
|
||||
"total": 21,
|
||||
"by_collection": {
|
||||
"jobs": 6,
|
||||
"papers": 6,
|
||||
"industry": 9
|
||||
},
|
||||
"by_status": {
|
||||
"jobs": {
|
||||
"active": 4,
|
||||
"unknown": 2
|
||||
},
|
||||
"papers": {
|
||||
"skimmed": 5,
|
||||
"queued": 1
|
||||
},
|
||||
"industry": {
|
||||
"analyzed": 8,
|
||||
"queued": 1
|
||||
}
|
||||
},
|
||||
"top_topics": [
|
||||
[
|
||||
"agent-evaluation",
|
||||
8
|
||||
],
|
||||
[
|
||||
"agent",
|
||||
6
|
||||
],
|
||||
[
|
||||
"memory",
|
||||
5
|
||||
],
|
||||
[
|
||||
"evaluation",
|
||||
4
|
||||
],
|
||||
[
|
||||
"coding-agent",
|
||||
4
|
||||
],
|
||||
[
|
||||
"agent-architecture",
|
||||
3
|
||||
],
|
||||
[
|
||||
"agent-safety",
|
||||
3
|
||||
],
|
||||
[
|
||||
"enterprise-ai",
|
||||
3
|
||||
],
|
||||
[
|
||||
"tool-use",
|
||||
2
|
||||
],
|
||||
[
|
||||
"workflow-agent",
|
||||
1
|
||||
],
|
||||
[
|
||||
"llm-infra",
|
||||
1
|
||||
],
|
||||
[
|
||||
"inference",
|
||||
1
|
||||
],
|
||||
[
|
||||
"harness",
|
||||
1
|
||||
],
|
||||
[
|
||||
"context-compression",
|
||||
1
|
||||
],
|
||||
[
|
||||
"agent-productization",
|
||||
1
|
||||
],
|
||||
[
|
||||
"agent-infra",
|
||||
1
|
||||
],
|
||||
[
|
||||
"agent-testing",
|
||||
1
|
||||
],
|
||||
[
|
||||
"harness-engineering",
|
||||
1
|
||||
],
|
||||
[
|
||||
"human-in-the-loop",
|
||||
1
|
||||
],
|
||||
[
|
||||
"gui-agent",
|
||||
1
|
||||
]
|
||||
],
|
||||
"top_skills": [
|
||||
[
|
||||
"agent",
|
||||
5
|
||||
],
|
||||
[
|
||||
"memory",
|
||||
4
|
||||
],
|
||||
[
|
||||
"multi-agent",
|
||||
4
|
||||
],
|
||||
[
|
||||
"agent-evaluation",
|
||||
4
|
||||
],
|
||||
[
|
||||
"planning",
|
||||
3
|
||||
],
|
||||
[
|
||||
"tool-use",
|
||||
3
|
||||
],
|
||||
[
|
||||
"rag",
|
||||
2
|
||||
],
|
||||
[
|
||||
"function-calling",
|
||||
2
|
||||
],
|
||||
[
|
||||
"inference",
|
||||
2
|
||||
],
|
||||
[
|
||||
"cloud",
|
||||
2
|
||||
],
|
||||
[
|
||||
"reasoning",
|
||||
1
|
||||
],
|
||||
[
|
||||
"state-management",
|
||||
1
|
||||
],
|
||||
[
|
||||
"serving",
|
||||
1
|
||||
],
|
||||
[
|
||||
"distributed-training",
|
||||
1
|
||||
],
|
||||
[
|
||||
"gpu",
|
||||
1
|
||||
],
|
||||
[
|
||||
"model-compression",
|
||||
1
|
||||
],
|
||||
[
|
||||
"context-engineering",
|
||||
1
|
||||
],
|
||||
[
|
||||
"llm-application",
|
||||
1
|
||||
],
|
||||
[
|
||||
"rl",
|
||||
1
|
||||
],
|
||||
[
|
||||
"coding-agent",
|
||||
1
|
||||
]
|
||||
],
|
||||
"top_companies": [
|
||||
[
|
||||
"百度",
|
||||
3
|
||||
],
|
||||
[
|
||||
"Microsoft",
|
||||
3
|
||||
],
|
||||
[
|
||||
"Google DeepMind",
|
||||
2
|
||||
],
|
||||
[
|
||||
"OpenAI",
|
||||
2
|
||||
],
|
||||
[
|
||||
"字节 Seed",
|
||||
1
|
||||
],
|
||||
[
|
||||
"DeepSeek",
|
||||
1
|
||||
],
|
||||
[
|
||||
"腾讯",
|
||||
1
|
||||
],
|
||||
[
|
||||
"ByteDance Seed",
|
||||
1
|
||||
],
|
||||
[
|
||||
"Alibaba Qwen",
|
||||
1
|
||||
]
|
||||
]
|
||||
}
|
||||
@@ -6,48 +6,59 @@ status: seed
|
||||
|
||||
## Current Snapshot
|
||||
|
||||
- date:
|
||||
- scope:
|
||||
- sources reviewed:
|
||||
- date: 2026-07-08
|
||||
- scope: large-company Agent work, engineering blogs, benchmarks, product/technical reports
|
||||
- sources reviewed: 9 starter items
|
||||
- source mix: official 9
|
||||
|
||||
## Strong Industry Signals
|
||||
|
||||
-
|
||||
- Enterprise Agent 正在从 demo 转向有权限、记忆、评估、trace、安全和平台入口的系统。
|
||||
- Computer use/GUI Agent 正在成为主流产品能力，但安全控制必须产品化。
|
||||
- Memory 的工业定义正在从“记住信息”转向“通过经验改善流程执行”。
|
||||
- Agent debug/eval 正在围绕 trajectory、constraint、failure taxonomy 和 benchmark 构建。
|
||||
- 大公司开始把 Agent adoption 当成组织生产力指标看，而不是单一模型能力。
|
||||
|
||||
## Weak or Marketing-heavy Signals
|
||||
|
||||
-
|
||||
- 单纯模型发布里提到 “agent frontier” 但缺少 benchmark、代码或系统细节时，只能作为方向信号。
|
||||
- Qwen-AgentWorld 当前抓取到的信息不足，需要补完整正文。
|
||||
- Agent usage growth 指标能说明采用趋势，但不能直接说明任务质量。
|
||||
|
||||
## Company Patterns
|
||||
|
||||
| Company | Agent Direction | Evidence | Notes |
|
||||
| --- | --- | --- | --- |
|
||||
| OpenAI | _TBD_ | _TBD_ | _TBD_ |
|
||||
| Anthropic | _TBD_ | _TBD_ | _TBD_ |
|
||||
| Google DeepMind | _TBD_ | _TBD_ | _TBD_ |
|
||||
| Microsoft | _TBD_ | _TBD_ | _TBD_ |
|
||||
| DeepSeek | _TBD_ | _TBD_ | _TBD_ |
|
||||
| 腾讯 | _TBD_ | _TBD_ | _TBD_ |
|
||||
| 字节 Seed | _TBD_ | _TBD_ | _TBD_ |
|
||||
| 阿里 Qwen | _TBD_ | _TBD_ | _TBD_ |
|
||||
| OpenAI | internal data agent, enterprise adoption, Codex as work agent | in-house data agent, agents transforming work | 系统细节集中在数据 Agent |
|
||||
| Anthropic | computer use and coding agents | not yet itemized | 下一轮补 Claude computer use / Claude Code |
|
||||
| Google DeepMind | computer use, agent security/control | Gemini computer use, AI Control Roadmap | 安全控制非常值得沉淀 |
|
||||
| Microsoft | memory benchmark, failure diagnosis, skill optimization | STATE-Bench, AgentRx, SkillOpt | 评估和工程方法密集 |
|
||||
| DeepSeek | Agent integration ecosystem and hiring wave | jobs item, awesome-deepseek-agent source candidate | 需补官方技术资料 |
|
||||
| 腾讯 | Agent testing / cloud harness signal | jobs item | 需补技术博客和北京岗位 |
|
||||
| 字节 Seed | productivity agent, GUI+MCP, coding agent | Seed2.1 official blog | 与岗位中的 harness/memory/context compression 呼应 |
|
||||
| 阿里 Qwen | AgentWorld / world models for general agents | Qwen-AgentWorld queued item | 需补正文和 Qwen3.x agent blogs |
|
||||
|
||||
## Implementation Patterns
|
||||
|
||||
- tool use:
|
||||
- memory:
|
||||
- evaluation:
|
||||
- safety:
|
||||
- deployment:
|
||||
- productization:
|
||||
- tool use: API/MCP/function calling 与 GUI/computer use 两条线并行。
|
||||
- memory: 从 raw history 转向 structured memory、stateful benchmark、experience consolidation。
|
||||
- evaluation: adoption metrics、benchmark、trajectory failure diagnosis、deterministic assertions。
|
||||
- safety: defense-in-depth、sandboxing、human confirmation、prompt injection guard、permission gating。
|
||||
- deployment: Slack/Web/IDE/CLI/MCP、enterprise platform、internal tools。
|
||||
- productization: coding agent、data agent、office/productivity agent、customer support/travel/shopping workflows。
|
||||
|
||||
## Follow-up Queue
|
||||
|
||||
-
|
||||
- Anthropic Claude computer use / Claude Code / model cards。
|
||||
- OpenAI workspace agents / Frontier / Codex enterprise coding posts。
|
||||
- Qwen AgentWorld and Qwen3.x agent blogs full extraction。
|
||||
- DeepSeek official GitHub agent ecosystem and talent site details。
|
||||
- Tencent AI Lab / Hunyuan / Yuanbao Agent technical posts。
|
||||
|
||||
## Links Back to Knowledge Base
|
||||
|
||||
- docs to update:
|
||||
- papers to read:
|
||||
- jobs to compare:
|
||||
- experiments to create:
|
||||
- projects affected:
|
||||
- docs to update: evaluation, observability, memory, safety, practice playbook
|
||||
- papers to read: OpenAgentSafety, memory survey, EvoMemBench, SWE-EVO, Dialogue-SWEBench
|
||||
- jobs to compare: Baidu Agent algorithm/fullstack, DeepSeek Agent hiring, Tencent Agent testing, ByteDance Seed Agent research
|
||||
- experiments to create: memory benchmark, trace/failure diagnosis, GUI+MCP workflow
|
||||
- projects affected: future knowledge frontend, coding-agent experiments, eval harness
|
||||
|
||||
@@ -0,0 +1,72 @@
|
||||
# Industry: ByteDance Seed - Seed2.1 Agent Productivity
|
||||
|
||||
---
|
||||
type: industry
|
||||
company: ByteDance Seed
|
||||
team: Seed
|
||||
title: "Seed2.1 Officially Released: Advancing AI Productivity"
|
||||
url: https://seed.bytedance.com/en/blog/seed2-1-officially-released-advancing-ai-productivity
|
||||
source_name: ByteDance Seed
|
||||
source_type: research-blog
|
||||
source_quality: official
|
||||
published_at:
|
||||
collected_at: 2026-07-08
|
||||
status: analyzed
|
||||
topics:
|
||||
- agent
|
||||
- coding-agent
|
||||
- gui-agent
|
||||
implementation_signals:
|
||||
- reinforcement-learning
|
||||
- gui-tool-use
|
||||
- mcp
|
||||
- crowdsourced-evaluation
|
||||
- internal-benchmark
|
||||
product_area:
|
||||
- productivity
|
||||
- coding-agent
|
||||
- office-agent
|
||||
models:
|
||||
- Seed2.1
|
||||
tools:
|
||||
- MCP
|
||||
benchmarks:
|
||||
- CreativeWork
|
||||
- ProgramBench
|
||||
related_papers:
|
||||
- 2026-pham-swe-evo
|
||||
related_jobs:
|
||||
- 2026-07-08-bytedance-seed-llm-agent-research-engineer
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
evidence_level: high
|
||||
relevance: high
|
||||
---
|
||||
|
||||
## One-line Takeaway
|
||||
|
||||
ByteDance Seed 把 Agent 能力放在办公生产力、GUI+MCP 工具协作和端到端 coding delivery 三个方向同时推进。
|
||||
|
||||
## What They Built or Claimed
|
||||
|
||||
Seed2.1 强调通用 Agent 和代码工程能力，使用 RL 提升 GUI/非 GUI 行动选择效率，并提出 CreativeWork 内部 benchmark 评估 Notion、Canva、Figma 等生产力场景中的 GUI 与 MCP 协同。
|
||||
|
||||
## Technical Signals
|
||||
|
||||
- architecture: model + GUI action space + MCP tools。
|
||||
- tool use: GUI 与 MCP 工具协调。
|
||||
- memory:
|
||||
- evaluation: CreativeWork、ProgramBench、众包开发者反馈、内部评估。
|
||||
- safety:
|
||||
- deployment: productivity 和 coding scenarios。
|
||||
- data: 真实代码仓库众包任务反馈。
|
||||
|
||||
## Evidence Quality
|
||||
|
||||
官方 Seed 博客，包含 benchmark、场景和评估方式，证据强；具体数据表需后续截图/原文复核。
|
||||
|
||||
## Gaps for Us
|
||||
|
||||
- 需要研究 GUI+MCP 的 Agent 工作流设计。
|
||||
@@ -0,0 +1,70 @@
|
||||
# Industry: Google DeepMind - AI Control Roadmap for Agents
|
||||
|
||||
---
|
||||
type: industry
|
||||
company: Google DeepMind
|
||||
team:
|
||||
title: Securing the future of AI agents
|
||||
url: https://deepmind.google/blog/securing-the-future-of-ai-agents/
|
||||
source_name: Google DeepMind
|
||||
source_type: technical-report
|
||||
source_quality: official
|
||||
published_at:
|
||||
collected_at: 2026-07-08
|
||||
status: analyzed
|
||||
topics:
|
||||
- agent-safety
|
||||
- agent-security
|
||||
- governance
|
||||
implementation_signals:
|
||||
- defense-in-depth
|
||||
- sandboxing
|
||||
- permission-control
|
||||
- prompt-injection-resistance
|
||||
product_area:
|
||||
- internal-agents
|
||||
- enterprise-agent
|
||||
models:
|
||||
-
|
||||
tools:
|
||||
-
|
||||
benchmarks:
|
||||
-
|
||||
related_papers:
|
||||
- 2025-vijayvargiya-openagentsafety
|
||||
- 2026-agent-safety-benchmark-taxonomy
|
||||
related_jobs:
|
||||
- 2026-07-08-tencent-cloud-ai-agent-test-engineer
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
evidence_level: high
|
||||
relevance: high
|
||||
---
|
||||
|
||||
## One-line Takeaway
|
||||
|
||||
Agent 安全需要把内部 Agent 当作潜在 insider threat 来设计权限、沙箱和逐步授权。
|
||||
|
||||
## What They Built or Claimed
|
||||
|
||||
Google DeepMind 介绍 AI Control Roadmap，用 defense-in-depth 保护 Google 内部系统，应对能力增强但不完美对齐的 Agent。
|
||||
|
||||
## Technical Signals
|
||||
|
||||
- architecture: alignment + system-level AI control + cybersecurity-style threat modeling。
|
||||
- tool use:
|
||||
- memory:
|
||||
- evaluation: 通过行为验证逐步授予权限。
|
||||
- safety: sandboxing、endpoint security、prompt injection resistance、权限分层。
|
||||
- deployment: 内部 Agent 系统治理。
|
||||
- data:
|
||||
|
||||
## Evidence Quality
|
||||
|
||||
官方技术/治理材料，适合作为 Agent 安全架构参考。
|
||||
|
||||
## Gaps for Us
|
||||
|
||||
- 需要把 Agent 权限分层写入实践 playbook。
|
||||
@@ -0,0 +1,70 @@
|
||||
# Industry: Google DeepMind - Computer Use in Gemini 3.5 Flash
|
||||
|
||||
---
|
||||
type: industry
|
||||
company: Google DeepMind
|
||||
team: Gemini
|
||||
title: Introducing computer use in Gemini 3.5 Flash
|
||||
url: https://deepmind.google/blog/introducing-computer-use-in-gemini-3-5-flash/
|
||||
source_name: Google DeepMind / Google Blog
|
||||
source_type: product-blog
|
||||
source_quality: official
|
||||
published_at: 2026-06-24
|
||||
collected_at: 2026-07-08
|
||||
status: analyzed
|
||||
topics:
|
||||
- computer-use
|
||||
- agent
|
||||
- tool-use
|
||||
implementation_signals:
|
||||
- browser-control
|
||||
- gui-control
|
||||
- safety
|
||||
- human-in-the-loop
|
||||
product_area:
|
||||
- enterprise-agent
|
||||
- automation
|
||||
models:
|
||||
- Gemini 3.5 Flash
|
||||
tools:
|
||||
- Gemini API
|
||||
- Gemini Enterprise Agent Platform
|
||||
benchmarks:
|
||||
-
|
||||
related_papers:
|
||||
- 2025-vijayvargiya-openagentsafety
|
||||
related_jobs:
|
||||
- 2026-07-08-baidu-aidu-agent-fullstack-engineer-beijing
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
evidence_level: high
|
||||
relevance: high
|
||||
---
|
||||
|
||||
## One-line Takeaway
|
||||
|
||||
Computer use 正在从独立模型能力并入主模型和企业 Agent 平台，安全控制成为产品能力的一部分。
|
||||
|
||||
## What They Built or Claimed
|
||||
|
||||
Google 表示 Gemini 3.5 Flash 原生支持 computer use，可让 Agent 在浏览器、移动端和桌面环境中看、推理并行动，支持长程和企业自动化任务。
|
||||
|
||||
## Technical Signals
|
||||
|
||||
- architecture: computer use 作为 Gemini 3.5 Flash 内置工具能力。
|
||||
- tool use: browser/mobile/desktop GUI actions。
|
||||
- memory:
|
||||
- evaluation: 提到 agentic computer use tasks，但需要进一步追具体 benchmark。
|
||||
- safety: 对 computer use 做 adversarial training；企业 safeguard 支持敏感/不可逆动作确认和间接 prompt injection 自动停止。
|
||||
- deployment: Gemini API 和 Gemini Enterprise Agent Platform。
|
||||
- data:
|
||||
|
||||
## Evidence Quality
|
||||
|
||||
官方产品/技术博客，有明确能力边界、安全控制和平台入口。
|
||||
|
||||
## Gaps for Us
|
||||
|
||||
- 需要整理 computer-use Agent 的安全 checklist。
|
||||
@@ -0,0 +1,68 @@
|
||||
# Industry: Microsoft Research - AgentRx
|
||||
|
||||
---
|
||||
type: industry
|
||||
company: Microsoft
|
||||
team: Microsoft Research
|
||||
title: "Systematic debugging for AI agents: Introducing the AgentRx framework"
|
||||
url: https://www.microsoft.com/en-us/research/blog/systematic-debugging-for-ai-agents-introducing-the-agentrx-framework/
|
||||
source_name: Microsoft Research
|
||||
source_type: research-blog
|
||||
source_quality: official
|
||||
published_at:
|
||||
collected_at: 2026-07-08
|
||||
status: analyzed
|
||||
topics:
|
||||
- agent-evaluation
|
||||
- observability
|
||||
- debugging
|
||||
implementation_signals:
|
||||
- trajectory-normalization
|
||||
- constraint-synthesis
|
||||
- constraint-checking
|
||||
- failure-taxonomy
|
||||
product_area:
|
||||
- agent-debugging
|
||||
models:
|
||||
-
|
||||
tools:
|
||||
- AgentRx
|
||||
benchmarks:
|
||||
- AgentRx Benchmark
|
||||
related_papers:
|
||||
- 2025-vijayvargiya-openagentsafety
|
||||
related_jobs:
|
||||
- 2026-07-08-tencent-cloud-ai-agent-test-engineer
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
evidence_level: high
|
||||
relevance: high
|
||||
---
|
||||
|
||||
## One-line Takeaway
|
||||
|
||||
Agent 调试需要从“最终失败”定位到 trajectory 中第一个不可恢复错误步骤。
|
||||
|
||||
## What They Built or Claimed
|
||||
|
||||
Microsoft Research 发布 AgentRx 和 115 条人工标注失败轨迹 benchmark，用于定位 Agent trajectory 的 critical failure step。
|
||||
|
||||
## Technical Signals
|
||||
|
||||
- architecture: trajectory normalization -> constraint synthesis -> guarded evaluation -> LLM judge。
|
||||
- tool use: 根据 tool schema 和 domain policy 生成可执行约束。
|
||||
- memory:
|
||||
- evaluation: 逐步检查约束，输出可审计验证日志。
|
||||
- safety: 区分计划偏离、幻觉、新信息发明等失败类型。
|
||||
- deployment: domain-agnostic debugging framework。
|
||||
- data: tau-bench、Flash、Magentic-One 失败轨迹。
|
||||
|
||||
## Evidence Quality
|
||||
|
||||
官方研究博客，包含框架、benchmark 规模、失败分类，证据强。
|
||||
|
||||
## Gaps for Us
|
||||
|
||||
- 本知识库需要设计统一 Agent trace schema 和失败 taxonomy。
|
||||
@@ -0,0 +1,68 @@
|
||||
# Industry: Microsoft Research - SkillOpt
|
||||
|
||||
---
|
||||
type: industry
|
||||
company: Microsoft
|
||||
team: Microsoft Research
|
||||
title: "SkillOpt: Agent skills as trainable parameters"
|
||||
url: https://www.microsoft.com/en-us/research/blog/skillopt-agent-skills-as-trainable-parameters/
|
||||
source_name: Microsoft Research
|
||||
source_type: research-blog
|
||||
source_quality: official
|
||||
published_at: 2026-06-30
|
||||
collected_at: 2026-07-08
|
||||
status: analyzed
|
||||
topics:
|
||||
- agent
|
||||
- prompt-optimization
|
||||
- skill-learning
|
||||
implementation_signals:
|
||||
- skill-files
|
||||
- validation-gating
|
||||
- bounded-edits
|
||||
- eval-loop
|
||||
product_area:
|
||||
- agent-framework
|
||||
models:
|
||||
-
|
||||
tools:
|
||||
-
|
||||
benchmarks:
|
||||
- six-benchmark-evaluation
|
||||
related_papers:
|
||||
-
|
||||
related_jobs:
|
||||
- 2026-07-08-deepseek-agent-hiring-wave-beijing-hangzhou
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
evidence_level: high
|
||||
relevance: high
|
||||
---
|
||||
|
||||
## One-line Takeaway
|
||||
|
||||
Agent skill/prompt 文件可以被当成外部可训练参数，用评估循环优化，而不是靠人工随手改。
|
||||
|
||||
## What They Built or Claimed
|
||||
|
||||
Microsoft Research 提出 SkillOpt，将 skill editing 转成训练过程，在不改模型权重的情况下提升 Agent 行为可靠性。
|
||||
|
||||
## Technical Signals
|
||||
|
||||
- architecture: frozen target model + external skill file optimization。
|
||||
- tool use:
|
||||
- memory:
|
||||
- evaluation: 六个 benchmark、七个目标模型、三种执行模式。
|
||||
- safety: bounded text edits、validation gating、rejected-edit feedback、slow/meta updates。
|
||||
- deployment: 可审计技能文件。
|
||||
- data:
|
||||
|
||||
## Evidence Quality
|
||||
|
||||
官方研究博客给出实验覆盖和机制细节，证据强。
|
||||
|
||||
## Gaps for Us
|
||||
|
||||
- 可以把本项目的 `skills`、prompt、playbook 也纳入 eval-driven 优化。
|
||||
@@ -0,0 +1,71 @@
|
||||
# Industry: Microsoft - STATE-Bench
|
||||
|
||||
---
|
||||
type: industry
|
||||
company: Microsoft
|
||||
team: Microsoft Open Source
|
||||
title: "Introducing STATE-Bench: a benchmark for AI agent memory"
|
||||
url: https://opensource.microsoft.com/blog/2026/05/19/introducing-state-bench-a-benchmark-for-ai-agent-memory/
|
||||
source_name: Microsoft Open Source Blog
|
||||
source_type: benchmark
|
||||
source_quality: official
|
||||
published_at: 2026-05-19
|
||||
collected_at: 2026-07-08
|
||||
status: analyzed
|
||||
topics:
|
||||
- memory
|
||||
- agent-evaluation
|
||||
- enterprise-ai
|
||||
implementation_signals:
|
||||
- stateful-environment
|
||||
- user-simulator
|
||||
- deterministic-assertions
|
||||
- bring-your-own-memory
|
||||
product_area:
|
||||
- customer-support
|
||||
- travel
|
||||
- shopping
|
||||
models:
|
||||
-
|
||||
tools:
|
||||
- STATE-Bench
|
||||
benchmarks:
|
||||
- STATE-Bench
|
||||
related_papers:
|
||||
- 2026-memory-agent-survey
|
||||
- 2026-wang-evomembench
|
||||
related_jobs:
|
||||
- 2026-07-08-baidu-aidu-agent-algorithm-engineer-beijing
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
evidence_level: high
|
||||
relevance: high
|
||||
---
|
||||
|
||||
## One-line Takeaway
|
||||
|
||||
Memory 评估应衡量 Agent 是否因经验而更会执行流程，而不是只测试远距离事实召回。
|
||||
|
||||
## What They Built or Claimed
|
||||
|
||||
Microsoft 发布 STATE-Bench，一个开源、memory-agnostic benchmark，用现实企业任务评估 Agent 是否能通过经验改进。
|
||||
|
||||
## Technical Signals
|
||||
|
||||
- architecture: 多轮对话 loop + stateful environment + deterministic assertions。
|
||||
- tool use: 任务涉及查询、校验策略、计算费用、确认并执行。
|
||||
- memory: bring-your-own-memory 接口。
|
||||
- evaluation: 450 个任务，覆盖 travel、customer support、shopping。
|
||||
- safety: 状态改变任务会带来真实成本，强调流程合规。
|
||||
- deployment: open-source benchmark。
|
||||
- data: 预填充数据库和用户模拟器。
|
||||
|
||||
## Evidence Quality
|
||||
|
||||
官方开源博客，包含任务数、领域、评估 loop 和 GitHub 入口，证据强。
|
||||
|
||||
## Gaps for Us
|
||||
|
||||
- 可以基于 STATE-Bench 思路做一个小型企业流程 Agent eval。
|
||||
@@ -0,0 +1,75 @@
|
||||
# Industry: OpenAI - How Agents Are Transforming Work
|
||||
|
||||
---
|
||||
type: industry
|
||||
company: OpenAI
|
||||
team:
|
||||
title: How agents are transforming work
|
||||
url: https://openai.com/index/how-agents-are-transforming-work/
|
||||
source_name: OpenAI
|
||||
source_type: research-blog
|
||||
source_quality: official
|
||||
published_at:
|
||||
collected_at: 2026-07-08
|
||||
status: analyzed
|
||||
topics:
|
||||
- agent
|
||||
- enterprise-ai
|
||||
- coding-agent
|
||||
implementation_signals:
|
||||
- adoption-metrics
|
||||
- workflow-change
|
||||
product_area:
|
||||
- coding-agent
|
||||
- enterprise-work
|
||||
models:
|
||||
- Codex
|
||||
tools:
|
||||
-
|
||||
benchmarks:
|
||||
-
|
||||
related_papers:
|
||||
-
|
||||
related_jobs:
|
||||
- 2026-07-08-deepseek-agent-hiring-wave-beijing-hangzhou
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
evidence_level: medium
|
||||
relevance: high
|
||||
---
|
||||
|
||||
## One-line Takeaway
|
||||
|
||||
Agent 采用正在从“聊天辅助”变成“可计量的 agentic labor”，企业内部不同职能的增长速度差异值得追踪。
|
||||
|
||||
## What They Built or Claimed
|
||||
|
||||
OpenAI 公开了内部 Codex 使用变化，显示过去六个月内研究、客服、工程、法务等部门的 agentic usage 明显增长，并认为员工正在从 chatbot 转向 agent 作为主要 AI 交互形态。
|
||||
|
||||
## Technical Signals
|
||||
|
||||
- architecture: 未详细公开。
|
||||
- tool use: Codex 作为工作代理。
|
||||
- memory:
|
||||
- evaluation: 使用 adoption / output token 增长作为组织层面的采用指标。
|
||||
- safety:
|
||||
- deployment: 企业内部多部门使用。
|
||||
- data: 部门级使用增长。
|
||||
|
||||
## Evidence Quality
|
||||
|
||||
官方数据型博客，适合看采用趋势；技术实现细节较少。
|
||||
|
||||
## Links
|
||||
|
||||
- papers:
|
||||
- jobs: DeepSeek Agent hiring wave
|
||||
- experiments:
|
||||
- projects:
|
||||
- docs: practice playbook
|
||||
|
||||
## Gaps for Us
|
||||
|
||||
- 需要设计自己的 Agent 使用指标：任务数、回合数、人工改动量、成功率、耗时和成本。
|
||||
@@ -0,0 +1,81 @@
|
||||
# Industry: OpenAI - In-house Data Agent
|
||||
|
||||
---
|
||||
type: industry
|
||||
company: OpenAI
|
||||
team: data / engineering
|
||||
title: Inside OpenAI's in-house data agent
|
||||
url: https://openai.com/index/inside-our-in-house-data-agent/
|
||||
source_name: OpenAI
|
||||
source_type: engineering-blog
|
||||
source_quality: official
|
||||
published_at:
|
||||
collected_at: 2026-07-08
|
||||
status: analyzed
|
||||
topics:
|
||||
- agent
|
||||
- data-agent
|
||||
- enterprise-ai
|
||||
implementation_signals:
|
||||
- memory
|
||||
- rag
|
||||
- mcp
|
||||
- eval
|
||||
- permissions
|
||||
product_area:
|
||||
- data-analysis
|
||||
- internal-tools
|
||||
models:
|
||||
- GPT-5.2
|
||||
- Codex
|
||||
tools:
|
||||
- Evals API
|
||||
- Embeddings API
|
||||
- MCP
|
||||
benchmarks:
|
||||
-
|
||||
related_papers:
|
||||
- 2026-memory-agent-survey
|
||||
related_jobs:
|
||||
- 2026-07-08-baidu-aidu-agent-fullstack-engineer-beijing
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
evidence_level: high
|
||||
relevance: high
|
||||
---
|
||||
|
||||
## One-line Takeaway
|
||||
|
||||
高价值内部 Agent 不是只接一个数据库，而是把权限、表级知识、人工注释、代码增强、组织知识、记忆和运行时上下文做成系统。
|
||||
|
||||
## What They Built or Claimed
|
||||
|
||||
OpenAI 构建了内部数据 Agent，用自然语言帮助员工探索和分析公司数据平台。资料描述了多层上下文、Agent API、Slack/Web/CLI/MCP 入口、记忆系统和安全边界。
|
||||
|
||||
## Technical Signals
|
||||
|
||||
- architecture: Agent API + internal knowledge + warehouse/platform sources + model/tool loop。
|
||||
- tool use: 查询数据、语义检索、精确文本检索、MCP 连接。
|
||||
- memory: 持续学习的记忆系统，用于提升后续查询。
|
||||
- evaluation: 使用 Evals API 作为构建工具之一。
|
||||
- safety: 内部权限和工作流是设计的一部分。
|
||||
- deployment: Slack、Web、IDE、Codex CLI、内部 ChatGPT。
|
||||
- data: 表使用、人工注释、Codex enrichment、institutional knowledge、runtime context。
|
||||
|
||||
## Evidence Quality
|
||||
|
||||
官方工程博客，有系统结构、上下文层和落地入口，证据强。
|
||||
|
||||
## Links
|
||||
|
||||
- papers: memory survey
|
||||
- jobs: Baidu Agent fullstack
|
||||
- experiments: data-agent-context-stack
|
||||
- projects:
|
||||
- docs: evaluation, memory, tool use
|
||||
|
||||
## Gaps for Us
|
||||
|
||||
- 需要沉淀一个“企业数据 Agent 架构”专题。
|
||||
@@ -0,0 +1,67 @@
|
||||
# Industry: Qwen - AgentWorld
|
||||
|
||||
---
|
||||
type: industry
|
||||
company: Alibaba Qwen
|
||||
team: Qwen
|
||||
title: "Qwen-AgentWorld: Language World Models for General Agents"
|
||||
url: https://qwen.ai/blog?id=qwen-agentworld
|
||||
source_name: Qwen Blog
|
||||
source_type: research-blog
|
||||
source_quality: official
|
||||
published_at:
|
||||
collected_at: 2026-07-08
|
||||
status: queued
|
||||
topics:
|
||||
- agent
|
||||
- world-model
|
||||
- evaluation
|
||||
implementation_signals:
|
||||
- simulated-environment
|
||||
- world-model
|
||||
- multi-domain-agent
|
||||
product_area:
|
||||
- general-agent
|
||||
models:
|
||||
- Qwen
|
||||
tools:
|
||||
-
|
||||
benchmarks:
|
||||
-
|
||||
related_papers:
|
||||
-
|
||||
related_jobs:
|
||||
-
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
evidence_level: medium
|
||||
relevance: high
|
||||
---
|
||||
|
||||
## One-line Takeaway
|
||||
|
||||
Qwen 正在把通用 Agent 能力和“语言世界模型/环境模拟”联系起来，值得后续重点读完整技术细节。
|
||||
|
||||
## What They Built or Claimed
|
||||
|
||||
搜索摘要显示 Qwen 发布 Qwen-AgentWorld，一个 native language world model，用于模拟七个领域的 Agent 环境。
|
||||
|
||||
## Technical Signals
|
||||
|
||||
- architecture: language world model for simulated agent environments。
|
||||
- tool use:
|
||||
- memory:
|
||||
- evaluation: 多领域环境模拟可能用于 Agent 训练或评估。
|
||||
- safety:
|
||||
- deployment:
|
||||
- data:
|
||||
|
||||
## Evidence Quality
|
||||
|
||||
官方博客入口存在，但当前抓取未展开正文；先作为 queued industry item，后续需要浏览器或脚本补全。
|
||||
|
||||
## Gaps for Us
|
||||
|
||||
- 需要详细阅读 Qwen-AgentWorld，判断它是训练环境、评估环境还是产品化组件。
|
||||
@@ -0,0 +1,95 @@
|
||||
# Job: 百度 - 2027AIDU 智能体算法工程师
|
||||
|
||||
---
|
||||
type: job
|
||||
company: 百度
|
||||
role: 2027AIDU-智能体算法工程师
|
||||
location: 北京市
|
||||
source_url: https://talent.baidu.com/jobs/list?projectType=3&recruitType=GRADUATE
|
||||
source_name: 百度校园招聘
|
||||
source_quality: official
|
||||
collected_at: 2026-07-08
|
||||
posted_at: 2026-05-12
|
||||
first_seen_at: 2026-07-08
|
||||
last_checked_at: 2026-07-08
|
||||
snapshot_type: new
|
||||
job_code: J99969
|
||||
previous_snapshot:
|
||||
status: active
|
||||
level: graduate
|
||||
team: AIDU项目
|
||||
business_area: search / conversation / office / data platform / entertainment
|
||||
employment_type: campus
|
||||
salary:
|
||||
skills:
|
||||
- agent
|
||||
- planning
|
||||
- tool-use
|
||||
- memory
|
||||
- multi-agent
|
||||
- rag
|
||||
- agent-evaluation
|
||||
topics:
|
||||
- agent-architecture
|
||||
- evaluation
|
||||
- memory
|
||||
models:
|
||||
-
|
||||
related_papers:
|
||||
- 2025-vijayvargiya-openagentsafety
|
||||
- 2026-memory-agent-survey
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
relevance: high
|
||||
---
|
||||
|
||||
## Source Snapshot
|
||||
|
||||
- source: 百度官方招聘页
|
||||
- collected_at: 2026-07-08
|
||||
- availability: 页面可访问，岗位列表显示发布日期 2026-05-12。
|
||||
|
||||
## Change Snapshot
|
||||
|
||||
- first_seen_at: 2026-07-08
|
||||
- last_checked_at: 2026-07-08
|
||||
- snapshot_type: new
|
||||
- previous_snapshot:
|
||||
- changed_fields:
|
||||
|
||||
## Raw JD Summary
|
||||
|
||||
百度 AIDU 的智能体算法岗位要求负责 AI Agent 设计和研发，覆盖感知、决策、执行闭环、多智能体协作、长期记忆、推理、规划、工具调用、反思、代码生成、RAG 融合和 Agent 评测体系建设。
|
||||
|
||||
## Responsibilities
|
||||
|
||||
- 设计和研发 AI Agent 的核心机制。
|
||||
- 推动 Agent 在搜索、对话、办公、数据平台、网盘文库、生活娱乐等产品场景落地。
|
||||
- 建设评测体系，优化成功率、延迟、成本和用户体验。
|
||||
|
||||
## Requirements
|
||||
|
||||
- 熟悉 ReAct、AutoGPT、CoT 等范式。
|
||||
- 熟悉 LangChain、AutoGen、CrewAI 等 Agent 框架。
|
||||
- 理解 RAG 与 Agent 融合架构。
|
||||
|
||||
## Extracted Signals
|
||||
|
||||
- hard requirements: planning、tool use、memory、multi-agent、RAG、eval。
|
||||
- preferred skills: Agent 框架实践、产品场景落地。
|
||||
- hidden signals: 百度把 Agent 从单点算法能力看成端到端系统能力。
|
||||
- business direction: 搜索、对话、办公、数据平台和泛生活娱乐场景。
|
||||
|
||||
## Knowledge Links
|
||||
|
||||
- concepts: Agent loop, tool use, memory, evaluation
|
||||
- papers: OpenAgentSafety, memory survey
|
||||
- experiments: eval harness, memory benchmark
|
||||
- projects:
|
||||
|
||||
## Gaps for Us
|
||||
|
||||
- 需要补充主流 Agent 框架的横向比较。
|
||||
- 需要建立 Agent 评测指标和任务集。
|
||||
@@ -0,0 +1,92 @@
|
||||
# Job: 百度 - 2027AIDU Agent 应用全栈工程师
|
||||
|
||||
---
|
||||
type: job
|
||||
company: 百度
|
||||
role: 2027AIDU-Agent应用全栈工程师
|
||||
location: 北京市
|
||||
source_url: https://talent.baidu.com/jobs/list?projectType=3&recruitType=GRADUATE
|
||||
source_name: 百度校园招聘
|
||||
source_quality: official
|
||||
collected_at: 2026-07-08
|
||||
posted_at: 2026-05-12
|
||||
first_seen_at: 2026-07-08
|
||||
last_checked_at: 2026-07-08
|
||||
snapshot_type: new
|
||||
job_code: J99974
|
||||
previous_snapshot:
|
||||
status: active
|
||||
level: graduate
|
||||
team: AIDU项目
|
||||
business_area: search / healthcare / enterprise service / data analysis / office automation
|
||||
employment_type: campus
|
||||
salary:
|
||||
skills:
|
||||
- agent
|
||||
- planning
|
||||
- tool-use
|
||||
- function-calling
|
||||
- memory
|
||||
- reasoning
|
||||
- state-management
|
||||
- multi-agent
|
||||
- rag
|
||||
- agent-evaluation
|
||||
topics:
|
||||
- workflow-agent
|
||||
- agent-architecture
|
||||
- evaluation
|
||||
models:
|
||||
-
|
||||
related_papers:
|
||||
- 2026-dialogue-swebench
|
||||
- 2026-swe-evo
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
relevance: high
|
||||
---
|
||||
|
||||
## Source Snapshot
|
||||
|
||||
- source: 百度官方招聘页
|
||||
- collected_at: 2026-07-08
|
||||
- availability: 页面可访问，岗位列表显示发布日期 2026-05-12。
|
||||
|
||||
## Change Snapshot
|
||||
|
||||
- first_seen_at: 2026-07-08
|
||||
- last_checked_at: 2026-07-08
|
||||
- snapshot_type: new
|
||||
- previous_snapshot:
|
||||
- changed_fields:
|
||||
|
||||
## Raw JD Summary
|
||||
|
||||
岗位强调构建下一代 Autonomous Agent 系统，要求从对话工具转向可执行任务的智能体，覆盖 planning、acting、reflection、Tool/API 调用、长短期记忆、多轮推理、状态管理、多 Agent 协作、RAG 融合和 Eval。
|
||||
|
||||
## Responsibilities
|
||||
|
||||
- 构建具备任务执行能力的 Agent 应用系统。
|
||||
- 搭建 Tool/API、memory、reasoning、state 等核心能力。
|
||||
- 推动 Agent 在搜索、医疗、企业服务、数据分析和自动化办公等场景落地。
|
||||
|
||||
## Extracted Signals
|
||||
|
||||
- hard requirements: full-stack Agent engineering, tool/function calling, memory, state, eval。
|
||||
- preferred skills: 前沿方向跟踪，包括 Tool Learning、Self-Improvement、World Model、多模态 Agent。
|
||||
- hidden signals: 全栈岗位已经不只是 Web 工程，开始要求 Agent runtime 和评估能力。
|
||||
- business direction: 企业服务、数据分析、自动化办公。
|
||||
|
||||
## Knowledge Links
|
||||
|
||||
- concepts: workflow-agent, human-in-the-loop, eval
|
||||
- papers: Dialogue-SWEBench, SWE-EVO
|
||||
- experiments: workflow-agent prototype
|
||||
- projects:
|
||||
|
||||
## Gaps for Us
|
||||
|
||||
- 需要补一个 Agent runtime / state management 的工程专题。
|
||||
- 需要整理 Tool Learning 和 Self-Improvement 的论文清单。
|
||||
@@ -0,0 +1,70 @@
|
||||
# Job: 百度 - 2027AIDU 大模型 Infra 工程师
|
||||
|
||||
---
|
||||
type: job
|
||||
company: 百度
|
||||
role: 2027AIDU-大模型Infra工程师
|
||||
location: 北京市
|
||||
source_url: https://talent.baidu.com/jobs/list?projectType=3&recruitType=GRADUATE
|
||||
source_name: 百度校园招聘
|
||||
source_quality: official
|
||||
collected_at: 2026-07-08
|
||||
posted_at: 2026-05-12
|
||||
first_seen_at: 2026-07-08
|
||||
last_checked_at: 2026-07-08
|
||||
snapshot_type: new
|
||||
job_code: J99967
|
||||
previous_snapshot:
|
||||
status: active
|
||||
level: graduate
|
||||
team: AIDU项目
|
||||
business_area: model infrastructure
|
||||
employment_type: campus
|
||||
salary:
|
||||
skills:
|
||||
- inference
|
||||
- serving
|
||||
- distributed-training
|
||||
- gpu
|
||||
- model-compression
|
||||
- cloud
|
||||
topics:
|
||||
- llm-infra
|
||||
- inference
|
||||
models:
|
||||
-
|
||||
related_papers:
|
||||
-
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
relevance: medium
|
||||
---
|
||||
|
||||
## Source Snapshot
|
||||
|
||||
- source: 百度官方招聘页
|
||||
- collected_at: 2026-07-08
|
||||
- availability: 页面可访问，岗位列表显示发布日期 2026-05-12。
|
||||
|
||||
## Raw JD Summary
|
||||
|
||||
大模型 Infra 岗位关注 AI 异构计算加速引擎、底层算子、通信库、编译系统、分布式系统和训练/推理效率优化。
|
||||
|
||||
## Extracted Signals
|
||||
|
||||
- hard requirements: inference engine、通信库、算子、编译优化、分布式系统。
|
||||
- hidden signals: Agent 能力落地会被成本、延迟和推理效率约束；JD 需求里 infra 和 Agent 应用是并行出现的。
|
||||
- business direction: 大模型平台化和成本优化。
|
||||
|
||||
## Knowledge Links
|
||||
|
||||
- concepts: inference, latency, cost control
|
||||
- papers:
|
||||
- experiments: agent cost/latency evaluation
|
||||
- projects:
|
||||
|
||||
## Gaps for Us
|
||||
|
||||
- Agent 知识库需要增加成本、延迟和推理服务专题。
|
||||
@@ -0,0 +1,72 @@
|
||||
# Job: 字节 Seed - 大模型 Agent 研究工程师
|
||||
|
||||
---
|
||||
type: job
|
||||
company: 字节 Seed
|
||||
role: 大模型Agent研究工程师-Seed
|
||||
location: unknown
|
||||
source_url: https://jobs.bytedance.com/experienced/position/7628902314323806469/detail
|
||||
source_name: 字节跳动招聘
|
||||
source_quality: official
|
||||
collected_at: 2026-07-08
|
||||
posted_at:
|
||||
first_seen_at: 2026-07-08
|
||||
last_checked_at: 2026-07-08
|
||||
snapshot_type: new
|
||||
job_code: 7628902314323806469
|
||||
previous_snapshot:
|
||||
status: unknown
|
||||
level: experienced
|
||||
team: Seed
|
||||
business_area: agent research and engineering
|
||||
employment_type: full-time
|
||||
salary:
|
||||
skills:
|
||||
- agent
|
||||
- memory
|
||||
- context-engineering
|
||||
- planning
|
||||
- multi-agent
|
||||
- llm-application
|
||||
topics:
|
||||
- harness
|
||||
- memory
|
||||
- context-compression
|
||||
models:
|
||||
- Seed
|
||||
related_papers:
|
||||
- 2026-memory-agent-survey
|
||||
- 2026-evomembench
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
relevance: high
|
||||
---
|
||||
|
||||
## Source Snapshot
|
||||
|
||||
- source: 字节跳动官方招聘页；网页正文由前端渲染，当前采集依赖搜索摘要。
|
||||
- collected_at: 2026-07-08
|
||||
- availability: URL 可定位到官方岗位页，但需要后续用浏览器或 API 抽取完整 JD。
|
||||
|
||||
## Raw JD Summary
|
||||
|
||||
搜索摘要显示该岗位负责大模型 Agent 系统研究与工程落地，围绕复杂任务执行持续优化 Agent 效果，关键词包括 harness、memory、context compression、planning/execution、multi-agent。
|
||||
|
||||
## Extracted Signals
|
||||
|
||||
- hard requirements: agent harness, memory, context compression, planning/execution, multi-agent。
|
||||
- hidden signals: Seed 把 Agent 作为研究与工程落地结合的问题，而不是纯应用开发。
|
||||
- business direction: 复杂任务执行和通用 Agent 能力。
|
||||
|
||||
## Knowledge Links
|
||||
|
||||
- concepts: context engineering, agent harness, memory
|
||||
- papers: Memory survey, EvoMemBench
|
||||
- experiments: context compression comparison
|
||||
- projects:
|
||||
|
||||
## Gaps for Us
|
||||
|
||||
- 需要后续用可重复脚本抓取字节招聘 JS 数据或手动浏览器快照。
|
||||
@@ -0,0 +1,78 @@
|
||||
# Job: DeepSeek - Agent 方向招聘组合
|
||||
|
||||
---
|
||||
type: job
|
||||
company: DeepSeek
|
||||
role: Agent 方向招聘组合
|
||||
location: 北京市 / 杭州市
|
||||
source_url: https://hub.baai.ac.cn/view/53416
|
||||
source_name: 智源社区转载量子位
|
||||
source_quality: secondary
|
||||
collected_at: 2026-07-08
|
||||
posted_at: 2026-03-27
|
||||
first_seen_at: 2026-07-08
|
||||
last_checked_at: 2026-07-08
|
||||
snapshot_type: new
|
||||
job_code:
|
||||
previous_snapshot:
|
||||
status: unknown
|
||||
level: mixed
|
||||
team: Agent / data evaluation / infra / product
|
||||
business_area: search / creation / multimodal / personal assistant / workflow
|
||||
employment_type: full-time / internship
|
||||
salary:
|
||||
skills:
|
||||
- agent
|
||||
- rl
|
||||
- agent-evaluation
|
||||
- tool-use
|
||||
- function-calling
|
||||
- memory
|
||||
- multi-agent
|
||||
- coding-agent
|
||||
- data-quality
|
||||
- inference
|
||||
topics:
|
||||
- agent-productization
|
||||
- evaluation
|
||||
- agent-infra
|
||||
models:
|
||||
- DeepSeek
|
||||
related_papers:
|
||||
- 2025-vijayvargiya-openagentsafety
|
||||
- 2026-agentrx
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
relevance: high
|
||||
---
|
||||
|
||||
## Source Snapshot
|
||||
|
||||
- source: 智源社区转载量子位文章，文中参考 DeepSeek 招聘入口。
|
||||
- collected_at: 2026-07-08
|
||||
- availability: 二手来源可访问；官方招聘站需要后续逐条核验。
|
||||
|
||||
## Raw JD Summary
|
||||
|
||||
该资料显示 DeepSeek 在 2026-03 开放一批 Agent 相关岗位，覆盖算法研究、数据评测、基础设施、产品经理和全栈工程。岗位信号集中在 RL/RLAIF、Agent 评测数据、工具调用、多轮交互、长期记忆、MCP、内部 Agent 框架、评测平台和 Agent runtime。
|
||||
|
||||
## Extracted Signals
|
||||
|
||||
- hard requirements: Agent algorithm, data/eval, infra, product scenario discovery。
|
||||
- preferred skills: 重度使用 Claude Code、Cursor、Copilot 等 AI 编程工具。
|
||||
- hidden signals: DeepSeek 的 Agent 招聘从基础模型研究转向产品化、数据闭环和运行时基础设施。
|
||||
- business direction: 搜索、内容创作、多模态、个人助理、自动化工作流。
|
||||
|
||||
## Knowledge Links
|
||||
|
||||
- concepts: coding-agent, agent evaluation, MCP, runtime isolation
|
||||
- papers: OpenAgentSafety, AgentRx
|
||||
- experiments: tool-use benchmark, eval data construction
|
||||
- projects:
|
||||
|
||||
## Gaps for Us
|
||||
|
||||
- 需要后续从 DeepSeek 官方招聘站逐条拆解岗位。
|
||||
- 需要补 MCP、Agent runtime isolation、AI coding tool 使用能力的实践笔记。
|
||||
@@ -0,0 +1,72 @@
|
||||
# Job: 腾讯云 - AI Agent 测试工程师
|
||||
|
||||
---
|
||||
type: job
|
||||
company: 腾讯
|
||||
role: 腾讯云-AI Agent测试工程师
|
||||
location: 深圳市
|
||||
source_url: https://careers.tencent.com/jobdesc.html?postId=2055555661177204736
|
||||
source_name: 腾讯招聘
|
||||
source_quality: official
|
||||
collected_at: 2026-07-08
|
||||
posted_at: 2026-06-03
|
||||
first_seen_at: 2026-07-08
|
||||
last_checked_at: 2026-07-08
|
||||
snapshot_type: new
|
||||
job_code: 2055555661177204736
|
||||
previous_snapshot:
|
||||
status: active
|
||||
level: experienced
|
||||
team: CSIG
|
||||
business_area: cloud agent testing
|
||||
employment_type: full-time
|
||||
salary:
|
||||
skills:
|
||||
- agent
|
||||
- agent-evaluation
|
||||
- observability
|
||||
- cloud
|
||||
- kubernetes
|
||||
- system-design
|
||||
topics:
|
||||
- agent-testing
|
||||
- harness-engineering
|
||||
models:
|
||||
-
|
||||
related_papers:
|
||||
- 2026-agentrx
|
||||
- 2025-vijayvargiya-openagentsafety
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
relevance: medium
|
||||
---
|
||||
|
||||
## Source Snapshot
|
||||
|
||||
- source: 腾讯官方招聘页；搜索结果显示岗位名、业务线、地点和更新时间。
|
||||
- collected_at: 2026-07-08
|
||||
- availability: 官方 URL 可定位，但正文需要后续浏览器或 API 抽取。
|
||||
- note: 非北京岗位，作为 Agent 测试/评估需求信号保留。
|
||||
|
||||
## Raw JD Summary
|
||||
|
||||
搜索摘要显示该岗位要求理解 LLM 原理与局限，具备大型项目的 AI Harness Engineering 或 Agent 落地工程实践，并熟悉云原生、分布式存储和云数据库等工程基础。
|
||||
|
||||
## Extracted Signals
|
||||
|
||||
- hard requirements: Agent 测试、harness engineering、云原生和分布式基础设施。
|
||||
- hidden signals: 腾讯云把 Agent 测试作为独立工程岗位，说明 Agent 评估和可靠性正在产品化。
|
||||
- business direction: 云上 Agent 服务、企业级测试和质量保障。
|
||||
|
||||
## Knowledge Links
|
||||
|
||||
- concepts: evaluation, observability, trace, harness
|
||||
- papers: AgentRx, OpenAgentSafety
|
||||
- experiments: failed trajectory debugging
|
||||
- projects:
|
||||
|
||||
## Gaps for Us
|
||||
|
||||
- 需要整理 Agent 测试工程和 harness engineering 的方法论。
|
||||
+33
-17
@@ -6,47 +6,63 @@ status: seed
|
||||
|
||||
## Current Snapshot
|
||||
|
||||
- date:
|
||||
- date: 2026-07-08
|
||||
- scope: 北京，大厂 Agent / LLM 相关岗位，滚动最近 6 个月
|
||||
- sample size:
|
||||
- source mix:
|
||||
- sample size: 6 starter items
|
||||
- source mix: official 5, secondary 1
|
||||
- caveat: 这是第一轮快速 sweep，不是完整市场统计；字节和腾讯部分页面需要后续用浏览器/API 补全正文。
|
||||
|
||||
## Requirement Changes
|
||||
|
||||
| Period | Signal | Companies | Evidence | Notes |
|
||||
| --- | --- | --- | --- | --- |
|
||||
| _TBD_ | _TBD_ | _TBD_ | _TBD_ | _TBD_ |
|
||||
| 2026-03 to 2026-07 | Agent 从研究走向产品化和工程化 | DeepSeek、百度、字节、腾讯 | DeepSeek 招聘波、百度 AIDU、字节 Seed、腾讯 Agent 测试 | 需要继续补齐官方岗位快照 |
|
||||
| 2026-05 to 2026-07 | Eval / harness / testing 变成显性能力 | 百度、腾讯、DeepSeek | 百度 J99969/J99974、腾讯 AI Agent 测试、DeepSeek 评测岗位 | 评估不再只是研究指标，而是岗位职责 |
|
||||
| 2026-05 to 2026-07 | Memory / context compression 进入核心要求 | 百度、字节 | 百度智能体算法、字节 Seed Agent 研究 | 与论文 memory benchmark 趋势重合 |
|
||||
|
||||
## High-frequency Skills
|
||||
|
||||
| Skill | Count | Companies | Notes |
|
||||
| --- | --- | --- | --- |
|
||||
| _TBD_ | _TBD_ | _TBD_ | _TBD_ |
|
||||
| agent | 5 | 百度、字节、DeepSeek、腾讯 | 多数岗位直接围绕 Agent 设计、落地或测试 |
|
||||
| memory | 4 | 百度、字节、DeepSeek | 长期记忆、短期记忆、context compression 反复出现 |
|
||||
| multi-agent | 4 | 百度、DeepSeek、字节 | 百度明确多 Agent 协作，DeepSeek 和字节出现复杂任务执行信号 |
|
||||
| agent-evaluation | 4 | 百度、DeepSeek、腾讯 | 评测体系、harness、测试工程成为硬需求 |
|
||||
| planning / tool-use | 3 | 百度、DeepSeek | planning、acting、tool/API/function calling 是核心机制 |
|
||||
| rag | 2 | 百度 | RAG+Agent 融合是应用岗位重点 |
|
||||
|
||||
## Company Differences
|
||||
|
||||
| Company | Signals | Notes |
|
||||
| --- | --- | --- |
|
||||
| DeepSeek | _TBD_ | _TBD_ |
|
||||
| 腾讯 | _TBD_ | _TBD_ |
|
||||
| 字节 | _TBD_ | _TBD_ |
|
||||
| 阿里 | _TBD_ | _TBD_ |
|
||||
| 百度 | _TBD_ | _TBD_ |
|
||||
| DeepSeek | Agent 产品化、数据评测、基础设施、AI coding tool 使用 | 当前主要来自二手文章和官方招聘入口线索，需逐条核验 |
|
||||
| 腾讯 | Agent 测试、harness engineering、云原生基础设施 | 当前保留一个非北京强信号，后续需补北京岗位 |
|
||||
| 字节 | Seed 侧重 harness、memory、context compression、planning/execution | 官方岗位页 JS 渲染，需要 API/浏览器补全 |
|
||||
| 阿里 | Qwen 行业实践已出现 AgentWorld/Qwen Agent 信号 | JD 尚未入库 |
|
||||
| 百度 | 岗位描述最完整，明确 Agent algorithm、fullstack、infra 三层 | 官方页面可直接抽取，适合作为第一批标杆 |
|
||||
|
||||
## Role Patterns
|
||||
|
||||
- Agent engineer:
|
||||
- LLM application engineer:
|
||||
- AI infra engineer:
|
||||
- AI product manager:
|
||||
- AI research engineer:
|
||||
- Agent engineer: planning、tool use、memory、multi-agent、RAG、eval。
|
||||
- LLM application engineer: 从对话应用转向可执行任务系统，要求 state management 和 tool/API。
|
||||
- AI infra engineer: 推理、通信库、算子、分布式系统和成本优化支撑 Agent 落地。
|
||||
- AI product manager: DeepSeek 资料显示 Agent 产品经理需要懂场景、训练数据和评测体系。
|
||||
- AI research engineer: 字节 Seed 强调 harness、memory、context compression 和复杂任务执行。
|
||||
|
||||
## Knowledge Gaps
|
||||
|
||||
这些是 JD 里出现但知识库尚未充分覆盖的方向：
|
||||
|
||||
-
|
||||
- Agent harness engineering
|
||||
- Agent runtime isolation and scheduling
|
||||
- Memory and context compression evaluation
|
||||
- RAG+Agent product architecture
|
||||
- AI coding tool proficiency as an engineering skill
|
||||
- Agent test / failure diagnosis methodology
|
||||
|
||||
## Follow-up Collection Queue
|
||||
|
||||
-
|
||||
- 补齐 DeepSeek 官方招聘站 Agent 相关岗位详情。
|
||||
- 抓取字节招聘 JS 页面或 API，补全 Seed Agent 岗位正文。
|
||||
- 继续找腾讯北京 Agent / 大模型评测 / 元宝 / 混元相关岗位。
|
||||
- 开始阿里 Qwen、百度社招、美团/快手/京东 Agent 岗位 sweep。
|
||||
|
||||
@@ -0,0 +1,64 @@
|
||||
# Paper: OpenAgentSafety
|
||||
|
||||
---
|
||||
type: paper
|
||||
title: OpenAgentSafety: A Comprehensive Framework for Evaluating Real-World AI Agent Safety
|
||||
authors: Sanidhya Vijayvargiya, Aditya Bharat Soni, Xuhui Zhou, Zora Zhiruo Wang, Nouha Dziri, Graham Neubig, Maarten Sap
|
||||
year: 2025
|
||||
venue: ICLR 2026 / IASEAI 2026
|
||||
url: https://arxiv.org/abs/2507.06134
|
||||
code_url:
|
||||
source: arxiv
|
||||
collected_at: 2026-07-08
|
||||
status: skimmed
|
||||
relevance: high
|
||||
topics:
|
||||
- agent-evaluation
|
||||
- agent-safety
|
||||
- tool-use
|
||||
methods:
|
||||
- multi-turn-agent-evaluation
|
||||
- rule-based-analysis
|
||||
- llm-as-judge
|
||||
benchmarks:
|
||||
- OpenAgentSafety
|
||||
models:
|
||||
-
|
||||
datasets:
|
||||
-
|
||||
related_concepts:
|
||||
- guardrail
|
||||
- human-in-the-loop
|
||||
related_jobs:
|
||||
- 2026-07-08-baidu-aidu-agent-algorithm-engineer-beijing
|
||||
- 2026-07-08-tencent-cloud-ai-agent-test-engineer
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
---
|
||||
|
||||
## One-line Takeaway
|
||||
|
||||
Agent safety 评估正在从简单拒答测试转向真实工具、多用户、多轮任务和可扩展评估框架。
|
||||
|
||||
## Problem
|
||||
|
||||
真实 Agent 会调用浏览器、文件系统、shell、消息平台和代码执行环境，传统安全 benchmark 对工具、任务和对抗场景覆盖不足。
|
||||
|
||||
## Core Idea
|
||||
|
||||
OpenAgentSafety 提供模块化框架，用真实工具和多轮任务评估 Agent 在八类风险下的行为，并结合规则分析和 LLM-as-judge。
|
||||
|
||||
## Evidence
|
||||
|
||||
论文报告覆盖 350+ 多轮、多用户任务，并评估五个 LLM 在 agentic 场景下的安全表现。
|
||||
|
||||
## Useful For Us
|
||||
|
||||
- 可作为 Agent eval harness 的安全维度参考。
|
||||
- 与 JD 中的 Agent 评测体系、成功率、稳定性、安全边界直接相关。
|
||||
|
||||
## Follow-up Experiments
|
||||
|
||||
- 选 10 个高风险工具任务，建立一个本地 mini OpenAgentSafety 风格测试集。
|
||||
@@ -0,0 +1,61 @@
|
||||
# Paper: Agent Safety Benchmark Taxonomy
|
||||
|
||||
---
|
||||
type: paper
|
||||
title: "Taxonomy and Consistency Analysis of Safety Benchmarks for AI Agents"
|
||||
authors:
|
||||
year: 2026
|
||||
venue:
|
||||
url: https://arxiv.org/html/2605.16282v1
|
||||
code_url:
|
||||
source: arxiv
|
||||
collected_at: 2026-07-08
|
||||
status: queued
|
||||
relevance: medium
|
||||
topics:
|
||||
- agent-safety
|
||||
- agent-evaluation
|
||||
methods:
|
||||
- benchmark-taxonomy
|
||||
- coverage-matrix
|
||||
benchmarks:
|
||||
- agent-safety-benchmarks
|
||||
models:
|
||||
-
|
||||
datasets:
|
||||
-
|
||||
related_concepts:
|
||||
- guardrail
|
||||
- red-teaming
|
||||
related_jobs:
|
||||
- 2026-07-08-tencent-cloud-ai-agent-test-engineer
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
---
|
||||
|
||||
## One-line Takeaway
|
||||
|
||||
Agent safety benchmark 还没有收敛，做评估时不能只依赖单一 benchmark。
|
||||
|
||||
## Problem
|
||||
|
||||
不同安全 benchmark 的风险覆盖、任务环境、评估方法和结论可能不一致。
|
||||
|
||||
## Core Idea
|
||||
|
||||
该论文整理 2023-2026 年的 behavioral agent-safety benchmarks，并用六轴 taxonomy 和 coverage matrix 分析方法选择如何影响结论。
|
||||
|
||||
## Evidence
|
||||
|
||||
搜索摘要显示论文 catalog 了 40 个 behavioral agent-safety benchmarks 和 5 个相邻 evaluator/defense/dataset artifact。
|
||||
|
||||
## Useful For Us
|
||||
|
||||
- 适合作为选择 Agent safety benchmark 的二级阅读材料。
|
||||
- 与 OpenAgentSafety、AgentRx、腾讯 Agent 测试岗位形成交叉验证。
|
||||
|
||||
## Follow-up Experiments
|
||||
|
||||
- 根据 taxonomy 给知识库建立自己的 Agent evaluation source map。
|
||||
@@ -0,0 +1,63 @@
|
||||
# Paper: Dialogue-SWEBench
|
||||
|
||||
---
|
||||
type: paper
|
||||
title: "Dialogue-SWEBench: A Benchmark for Dialogue-Driven Coding Agents"
|
||||
authors: Brendan King, Jeffrey Flanigan
|
||||
year: 2026
|
||||
venue:
|
||||
url: https://arxiv.org/html/2606.13995v1
|
||||
code_url: https://jlab-nlp.github.io/dialogue-swe-bench/
|
||||
source: arxiv
|
||||
collected_at: 2026-07-08
|
||||
status: skimmed
|
||||
relevance: high
|
||||
topics:
|
||||
- coding-agent
|
||||
- human-in-the-loop
|
||||
- agent-evaluation
|
||||
methods:
|
||||
- user-simulator
|
||||
- dialogue-benchmark
|
||||
- schema-guided-agent
|
||||
benchmarks:
|
||||
- Dialogue-SWEBench
|
||||
models:
|
||||
-
|
||||
datasets:
|
||||
- SWE-Bench Verified
|
||||
related_concepts:
|
||||
- human-in-the-loop
|
||||
- coding-agent
|
||||
related_jobs:
|
||||
- 2026-07-08-baidu-aidu-agent-fullstack-engineer-beijing
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
---
|
||||
|
||||
## One-line Takeaway
|
||||
|
||||
Coding agent 的真实能力不能只看自动修复，还要看它能不能通过对话澄清需求。
|
||||
|
||||
## Problem
|
||||
|
||||
现实中的 coding assistant 经常和用户交互，但主流 benchmark 多把它们评估为全自动系统。
|
||||
|
||||
## Core Idea
|
||||
|
||||
Dialogue-SWEBench 使用 persona-grounded user simulator，把 SWE task 改造成多轮对话评估，并提出 schema-guided agent 改善对话能力。
|
||||
|
||||
## Evidence
|
||||
|
||||
论文报告 schema-guided agent 相比强 baseline 提升 3-14%，并指出更强 coding model 不一定是更强 dialogue agent。
|
||||
|
||||
## Useful For Us
|
||||
|
||||
- 对“Agent 何时应该提问”很有启发。
|
||||
- 可以作为 coding-agent / human-in-the-loop 评估维度。
|
||||
|
||||
## Follow-up Experiments
|
||||
|
||||
- 给本知识库的 coding-agent 项目设计“澄清问题质量”评估。
|
||||
@@ -0,0 +1,65 @@
|
||||
# Paper: Memory for Autonomous LLM Agents
|
||||
|
||||
---
|
||||
type: paper
|
||||
title: "Memory for Autonomous LLM Agents: Mechanisms, Evaluation, and Emerging Frontiers"
|
||||
authors:
|
||||
year: 2026
|
||||
venue:
|
||||
url: https://arxiv.org/html/2603.07670v1
|
||||
code_url:
|
||||
source: arxiv
|
||||
collected_at: 2026-07-08
|
||||
status: skimmed
|
||||
relevance: high
|
||||
topics:
|
||||
- memory
|
||||
- agent-architecture
|
||||
- agent-evaluation
|
||||
methods:
|
||||
- write-manage-read-loop
|
||||
- retrieval-augmented-memory
|
||||
- reflective-memory
|
||||
- hierarchical-context
|
||||
benchmarks:
|
||||
-
|
||||
models:
|
||||
-
|
||||
datasets:
|
||||
-
|
||||
related_concepts:
|
||||
- memory
|
||||
- context-engineering
|
||||
related_jobs:
|
||||
- 2026-07-08-baidu-aidu-agent-algorithm-engineer-beijing
|
||||
- 2026-07-08-bytedance-seed-llm-agent-research-engineer
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
---
|
||||
|
||||
## One-line Takeaway
|
||||
|
||||
Memory 应该作为 Agent 系统的一等工程组件，而不是 prompt 里临时塞历史记录。
|
||||
|
||||
## Problem
|
||||
|
||||
长任务、多 session 和个性化场景中，单一上下文窗口无法保存并正确复用历史经验。
|
||||
|
||||
## Core Idea
|
||||
|
||||
该综述把 Agent memory 形式化为 write-manage-read loop，并从时间范围、表示基底和控制策略三个维度整理机制。
|
||||
|
||||
## Evidence
|
||||
|
||||
论文覆盖 2022 到 2026 年初的 memory 机制和评估进展，并总结从静态 recall 到多 session agentic benchmark 的转变。
|
||||
|
||||
## Useful For Us
|
||||
|
||||
- 用来设计知识库自己的 memory / context 管理专题。
|
||||
- 与百度和字节 JD 中的长期记忆、context compression、agent harness 强相关。
|
||||
|
||||
## Follow-up Experiments
|
||||
|
||||
- 用同一组任务比较 raw history、summary memory、retrieval memory 和 procedural memory。
|
||||
@@ -0,0 +1,65 @@
|
||||
# Paper: SWE-EVO
|
||||
|
||||
---
|
||||
type: paper
|
||||
title: "SWE-EVO: Benchmarking Coding Agents in Long-Horizon Software Evolution Scenarios"
|
||||
authors: Minh Vu Thai Pham, Tue Le, Dung Nguyen Manh, Huy Nhat Phan, Nghi D. Q. Bui
|
||||
year: 2026
|
||||
venue:
|
||||
url: https://arxiv.org/html/2512.18470v5
|
||||
code_url: https://github.com/SWE-EVO/SWE-EVO
|
||||
source: arxiv
|
||||
collected_at: 2026-07-08
|
||||
status: skimmed
|
||||
relevance: high
|
||||
topics:
|
||||
- coding-agent
|
||||
- agent-evaluation
|
||||
methods:
|
||||
- long-horizon-software-evolution
|
||||
- benchmark
|
||||
benchmarks:
|
||||
- SWE-EVO
|
||||
models:
|
||||
- OpenAI
|
||||
- DeepSeek
|
||||
- Zhipu
|
||||
- Qwen
|
||||
- Moonshot
|
||||
datasets:
|
||||
- release-notes
|
||||
related_concepts:
|
||||
- coding-agent
|
||||
- long-horizon-task
|
||||
related_jobs:
|
||||
- 2026-07-08-deepseek-agent-hiring-wave-beijing-hangzhou
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
---
|
||||
|
||||
## One-line Takeaway
|
||||
|
||||
Coding agent 评估正在从单 issue 修 bug 转向多文件、多 PR、版本演进式长任务。
|
||||
|
||||
## Problem
|
||||
|
||||
SWE-Bench 类任务偏单点 issue，而真实工程经常要求根据 release notes 进行系统性版本演进。
|
||||
|
||||
## Core Idea
|
||||
|
||||
SWE-EVO 从成熟 Python 项目的 release notes 构造 48 个长程任务，平均涉及 21 个文件，并提出 Fix Rate 捕捉部分进展。
|
||||
|
||||
## Evidence
|
||||
|
||||
论文显示强模型在 SWE-EVO 上的解决率远低于 SWE-Bench Verified，失败更多来自语义理解和长期指令跟随，而不是简单工具操作。
|
||||
|
||||
## Useful For Us
|
||||
|
||||
- 适合指导 coding-agent 项目的内部评测集设计。
|
||||
- 和 DeepSeek JD 中的 AI coding tool 使用、Agent infra、评测闭环相关。
|
||||
|
||||
## Follow-up Experiments
|
||||
|
||||
- 选一个开源 Python 项目，用 release notes 构造小规模 SWE-EVO 风格任务。
|
||||
@@ -0,0 +1,62 @@
|
||||
# Paper: EvoMemBench
|
||||
|
||||
---
|
||||
type: paper
|
||||
title: "EvoMemBench: Benchmarking Agent Memory from a Self-Evolving Perspective"
|
||||
authors: Yuyao Wang, Zhongjian Zhang, Mo Chi, Kaichi Yu, Yuhan Li, Miao Peng, Bing Tong, Chen Zhang, Yan Zhou, Jia Li
|
||||
year: 2026
|
||||
venue:
|
||||
url: https://arxiv.org/html/2605.18421
|
||||
code_url:
|
||||
source: arxiv
|
||||
collected_at: 2026-07-08
|
||||
status: skimmed
|
||||
relevance: high
|
||||
topics:
|
||||
- memory
|
||||
- agent-evaluation
|
||||
methods:
|
||||
- memory-benchmark
|
||||
- self-evolving-agent
|
||||
benchmarks:
|
||||
- EvoMemBench
|
||||
models:
|
||||
-
|
||||
datasets:
|
||||
-
|
||||
related_concepts:
|
||||
- memory
|
||||
- procedural-memory
|
||||
related_jobs:
|
||||
- 2026-07-08-baidu-aidu-agent-algorithm-engineer-beijing
|
||||
- 2026-07-08-bytedance-seed-llm-agent-research-engineer
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
---
|
||||
|
||||
## One-line Takeaway
|
||||
|
||||
Memory 的价值取决于任务类型和上下文约束；长上下文 baseline 仍然很强，不能假设外部记忆天然更好。
|
||||
|
||||
## Problem
|
||||
|
||||
现有 Agent benchmark 主要评估 reasoning、planning、execution，对 memory 的更新、修正和复用能力覆盖不足。
|
||||
|
||||
## Core Idea
|
||||
|
||||
EvoMemBench 按 memory scope 和 memory content 两条轴设计 benchmark，比较 15 种 memory 方法和长上下文 baseline。
|
||||
|
||||
## Evidence
|
||||
|
||||
结果显示 memory 在当前上下文不足或任务较难时更有价值；知识密集任务更适合 retrieval memory，执行导向任务更依赖 procedural/long-term memory。
|
||||
|
||||
## Useful For Us
|
||||
|
||||
- 设计 memory 实验时要加入长上下文 baseline。
|
||||
- 不同岗位里提到 memory 时，要进一步区分事实记忆、过程记忆和跨任务经验。
|
||||
|
||||
## Follow-up Experiments
|
||||
|
||||
- 用 Agent 项目任务集做 memory-vs-long-context 对照实验。
|
||||
+27
-12
@@ -6,35 +6,50 @@ status: seed
|
||||
|
||||
## Current Snapshot
|
||||
|
||||
- date:
|
||||
- scope:
|
||||
- papers reviewed:
|
||||
- date: 2026-07-08
|
||||
- scope: Agent evaluation, memory, coding-agent benchmarks
|
||||
- papers reviewed: 6 starter items
|
||||
- caveat: sources are provisional; source approval workflow still needs user confirmation.
|
||||
|
||||
## Strong Signals
|
||||
|
||||
-
|
||||
- Agent safety/eval 正在转向真实工具、多轮任务、trajectory、失败定位和系统级 guardrail。
|
||||
- Memory 不能只看事实召回；更重要的是是否改善流程执行和跨任务决策。
|
||||
- Coding-agent benchmark 正在从单 issue patch 转向长程软件演进和人机对话。
|
||||
- JD 高频词和论文趋势高度重合：memory、eval、planning/tool use、coding agent。
|
||||
|
||||
## Weak or Unproven Claims
|
||||
|
||||
-
|
||||
- “加 memory 就更强”不成立；EvoMemBench 显示长上下文 baseline 很强，memory 价值依赖任务。
|
||||
- 单一 safety benchmark 不足以说明 Agent 安全；不同 benchmark 的方法差异可能改变结论。
|
||||
- SWE-Bench 类单点任务不足以代表真实 coding agent 生产力。
|
||||
|
||||
## Mature Directions
|
||||
|
||||
-
|
||||
- Agent evaluation harness
|
||||
- Memory benchmark and memory architecture comparison
|
||||
- Coding agent benchmark beyond SWE-Bench
|
||||
- Agent failure diagnosis and trajectory analysis
|
||||
|
||||
## Emerging Directions
|
||||
|
||||
-
|
||||
- Dialogue-driven coding agents
|
||||
- Stateful enterprise task benchmarks
|
||||
- Agent safety benchmark taxonomy
|
||||
- Skill/prompt optimization as trainable artifacts
|
||||
|
||||
## Experiment Candidates
|
||||
|
||||
| Idea | Related Papers | Metric | Notes |
|
||||
| --- | --- | --- | --- |
|
||||
| _TBD_ | _TBD_ | _TBD_ | _TBD_ |
|
||||
| Mini Agent Safety Suite | OpenAgentSafety, Agent safety benchmark taxonomy | unsafe action rate, refusal precision, tool misuse | 选 10-20 个本地工具任务先跑 |
|
||||
| Memory vs Long Context | Memory survey, EvoMemBench, STATE-Bench | success rate, token cost, latency, contradiction rate | 必须包含 long-context baseline |
|
||||
| Long-horizon Coding Eval | SWE-EVO, Dialogue-SWEBench | task success, fix rate, clarification quality | 从真实 repo release notes 构造小任务 |
|
||||
| Failure Step Diagnosis | AgentRx, OpenAgentSafety | critical failure step accuracy | 需要统一 trace schema |
|
||||
|
||||
## Links Back to Knowledge Base
|
||||
|
||||
- docs to update:
|
||||
- experiments to create:
|
||||
- projects affected:
|
||||
- job skills connected:
|
||||
- docs to update: evaluation, memory, coding-agent, safety
|
||||
- experiments to create: memory comparison, mini safety suite, coding-agent eval
|
||||
- projects affected: future front-end/indexing project, coding-agent experiments
|
||||
- job skills connected: agent-evaluation, memory, coding-agent, tool-use, planning
|
||||
|
||||
@@ -4,7 +4,12 @@ status: seed
|
||||
|
||||
| Priority | Paper | Topic | Reason | Status | Link |
|
||||
| --- | --- | --- | --- | --- | --- |
|
||||
| _TBD_ | _TBD_ | _TBD_ | _TBD_ | queued | _TBD_ |
|
||||
| P0 | OpenAgentSafety | agent safety / evaluation | 真实工具、多轮、多用户安全评估，直接对应 Agent eval 需求 | skimmed | https://arxiv.org/abs/2507.06134 |
|
||||
| P0 | Memory for Autonomous LLM Agents | memory | 综述 memory 机制、评估和工程挑战，对 JD 高频 memory/context 信号有解释力 | skimmed | https://arxiv.org/html/2603.07670v1 |
|
||||
| P0 | EvoMemBench | memory benchmark | 说明 memory 不总是优于长上下文，适合作为实验设计依据 | skimmed | https://arxiv.org/html/2605.18421 |
|
||||
| P0 | SWE-EVO | coding agent benchmark | 长程软件演进任务比 SWE-Bench 更接近真实工程 | skimmed | https://arxiv.org/html/2512.18470v5 |
|
||||
| P1 | Dialogue-SWEBench | human-in-the-loop coding agent | 评估 coding agent 澄清需求能力，补足全自动 benchmark 盲区 | skimmed | https://arxiv.org/html/2606.13995v1 |
|
||||
| P1 | Taxonomy and Consistency Analysis of Safety Benchmarks for AI Agents | safety benchmark taxonomy | 用于选择和比较 Agent safety benchmark | queued | https://arxiv.org/html/2605.16282v1 |
|
||||
|
||||
## Priority Rules
|
||||
|
||||
|
||||
@@ -8,13 +8,13 @@
|
||||
|
||||
| Source | Status | Notes |
|
||||
| --- | --- | --- |
|
||||
| arXiv | candidate | 新论文发现和预印本 |
|
||||
| arXiv | provisional-used | 新论文发现和预印本；first sweep 已使用，待用户正式确认 |
|
||||
| OpenReview | candidate | ICLR、NeurIPS 等投稿和评审线索 |
|
||||
| ACL Anthology | candidate | NLP、RAG、tool use、evaluation 相关论文 |
|
||||
| Conference proceedings | candidate | NeurIPS、ICLR、ICML、ACL、EMNLP、KDD、WWW 等 |
|
||||
| Papers with Code | candidate | benchmark、代码和任务关联 |
|
||||
| Official project GitHub | candidate | 复现代码、模型卡和实验配置 |
|
||||
| Company research blogs | candidate | 只收技术细节足够强的研究内容 |
|
||||
| Official project GitHub | provisional-used | 复现代码、模型卡和实验配置；first sweep 已作为 code_url 使用 |
|
||||
| Company research blogs | provisional-used | 只收技术细节足够强的研究内容；first sweep 已在 industry 中使用 |
|
||||
|
||||
## Source Types
|
||||
|
||||
|
||||
@@ -0,0 +1,53 @@
|
||||
# Collection Tools
|
||||
|
||||
这些脚本用于把 JD、论文和行业资料从 Markdown 笔记变成可检索、可检查、可视化的数据。
|
||||
|
||||
## Scripts
|
||||
|
||||
### `new_item.py`
|
||||
|
||||
生成一个带 frontmatter 的新条目草稿。
|
||||
|
||||
```bash
|
||||
python3 tools/collection/new_item.py jobs "baidu aidu agent algorithm engineer beijing" \
|
||||
--title "Baidu AIDU Agent Algorithm Engineer" \
|
||||
--url "https://talent.baidu.com/jobs/list?projectType=3&recruitType=GRADUATE"
|
||||
```
|
||||
|
||||
支持的 collection:
|
||||
|
||||
- `jobs`
|
||||
- `papers`
|
||||
- `industry`
|
||||
|
||||
### `build_index.py`
|
||||
|
||||
扫描 `jobs/items/`、`papers/items/`、`industry/items/` 的 frontmatter，生成结构化索引。
|
||||
|
||||
```bash
|
||||
python3 tools/collection/build_index.py
|
||||
```
|
||||
|
||||
输出：
|
||||
|
||||
- `data/index.json`
|
||||
- `data/summary.json`
|
||||
|
||||
### `check_urls.py`
|
||||
|
||||
检查条目中的 `url`、`source_url`、`code_url` 是否可访问。
|
||||
|
||||
```bash
|
||||
python3 tools/collection/check_urls.py
|
||||
```
|
||||
|
||||
可加 `--json` 输出机器可读结果。
|
||||
|
||||
## Method
|
||||
|
||||
1. 搜索或导入资料。
|
||||
2. 用 `new_item.py` 生成草稿。
|
||||
3. 人工整理摘要、标签、判断和关联。
|
||||
4. 运行 `build_index.py` 更新结构化数据。
|
||||
5. 运行 `check_urls.py` 检查来源可达性。
|
||||
6. 更新对应 insight 文件。
|
||||
Binary file not shown.
Executable
+117
@@ -0,0 +1,117 @@
|
||||
#!/usr/bin/env python3
|
||||
"""Build a JSON index from Markdown frontmatter."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import json
|
||||
from collections import Counter, defaultdict
|
||||
from pathlib import Path
|
||||
|
||||
|
||||
ROOT = Path(__file__).resolve().parents[2]
|
||||
COLLECTIONS = ("jobs", "papers", "industry")
|
||||
|
||||
|
||||
def parse_scalar(value: str):
|
||||
value = value.strip()
|
||||
if value == "":
|
||||
return None
|
||||
return value
|
||||
|
||||
|
||||
def parse_frontmatter(text: str) -> dict:
|
||||
start = text.find("\n---\n")
|
||||
if text.startswith("---\n"):
|
||||
start = -1
|
||||
body_start = 4
|
||||
elif start != -1:
|
||||
body_start = start + 5
|
||||
else:
|
||||
return {}
|
||||
end = text.find("\n---", body_start)
|
||||
if end == -1:
|
||||
return {}
|
||||
lines = text[body_start:end].splitlines()
|
||||
data: dict[str, object] = {}
|
||||
current_key: str | None = None
|
||||
for raw in lines:
|
||||
line = raw.rstrip()
|
||||
if not line:
|
||||
continue
|
||||
if line.startswith(" -"):
|
||||
if current_key:
|
||||
item = line[3:].strip()
|
||||
if item:
|
||||
data.setdefault(current_key, [])
|
||||
assert isinstance(data[current_key], list)
|
||||
data[current_key].append(item)
|
||||
continue
|
||||
if ":" in line:
|
||||
key, value = line.split(":", 1)
|
||||
key = key.strip()
|
||||
value = value.strip()
|
||||
current_key = key
|
||||
data[key] = [] if value == "" else parse_scalar(value)
|
||||
return data
|
||||
|
||||
|
||||
def iter_items():
|
||||
for collection in COLLECTIONS:
|
||||
item_dir = ROOT / collection / "items"
|
||||
for path in sorted(item_dir.glob("*.md")):
|
||||
if path.name == "README.md":
|
||||
continue
|
||||
text = path.read_text(encoding="utf-8")
|
||||
meta = parse_frontmatter(text)
|
||||
yield {
|
||||
"collection": collection,
|
||||
"path": str(path.relative_to(ROOT)),
|
||||
"title": meta.get("title") or meta.get("role") or path.stem,
|
||||
"type": meta.get("type") or collection.rstrip("s"),
|
||||
"meta": meta,
|
||||
}
|
||||
|
||||
|
||||
def main() -> int:
|
||||
items = list(iter_items())
|
||||
data_dir = ROOT / "data"
|
||||
data_dir.mkdir(exist_ok=True)
|
||||
(data_dir / "index.json").write_text(
|
||||
json.dumps(items, ensure_ascii=False, indent=2) + "\n",
|
||||
encoding="utf-8",
|
||||
)
|
||||
|
||||
by_collection = Counter(item["collection"] for item in items)
|
||||
by_status = defaultdict(Counter)
|
||||
topics = Counter()
|
||||
skills = Counter()
|
||||
companies = Counter()
|
||||
for item in items:
|
||||
meta = item["meta"]
|
||||
by_status[item["collection"]][meta.get("status") or "unknown"] += 1
|
||||
for topic in meta.get("topics") or []:
|
||||
topics[topic] += 1
|
||||
for skill in meta.get("skills") or []:
|
||||
skills[skill] += 1
|
||||
company = meta.get("company")
|
||||
if company:
|
||||
companies[str(company)] += 1
|
||||
|
||||
summary = {
|
||||
"total": len(items),
|
||||
"by_collection": dict(by_collection),
|
||||
"by_status": {key: dict(value) for key, value in by_status.items()},
|
||||
"top_topics": topics.most_common(20),
|
||||
"top_skills": skills.most_common(20),
|
||||
"top_companies": companies.most_common(20),
|
||||
}
|
||||
(data_dir / "summary.json").write_text(
|
||||
json.dumps(summary, ensure_ascii=False, indent=2) + "\n",
|
||||
encoding="utf-8",
|
||||
)
|
||||
print(json.dumps(summary, ensure_ascii=False, indent=2))
|
||||
return 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
raise SystemExit(main())
|
||||
Executable
+86
@@ -0,0 +1,86 @@
|
||||
#!/usr/bin/env python3
|
||||
"""Check source URLs stored in collection item frontmatter."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import json
|
||||
import urllib.error
|
||||
import urllib.request
|
||||
from pathlib import Path
|
||||
|
||||
from build_index import parse_frontmatter
|
||||
|
||||
|
||||
ROOT = Path(__file__).resolve().parents[2]
|
||||
URL_KEYS = ("url", "source_url", "code_url")
|
||||
|
||||
|
||||
def check_url(url: str, timeout: int) -> dict:
|
||||
request = urllib.request.Request(url, method="HEAD", headers={"User-Agent": "agent-kb-url-check/1.0"})
|
||||
try:
|
||||
with urllib.request.urlopen(request, timeout=timeout) as response:
|
||||
return {"url": url, "ok": True, "status": response.status}
|
||||
except urllib.error.HTTPError as exc:
|
||||
if exc.code in (403, 405):
|
||||
return check_url_get(url, timeout, exc.code)
|
||||
return {"url": url, "ok": False, "status": exc.code, "error": str(exc)}
|
||||
except Exception as exc: # noqa: BLE001
|
||||
return {"url": url, "ok": False, "status": None, "error": str(exc)}
|
||||
|
||||
|
||||
def check_url_get(url: str, timeout: int, head_status: int) -> dict:
|
||||
request = urllib.request.Request(url, method="GET", headers={"User-Agent": "agent-kb-url-check/1.0"})
|
||||
try:
|
||||
with urllib.request.urlopen(request, timeout=timeout) as response:
|
||||
return {"url": url, "ok": True, "status": response.status, "head_status": head_status}
|
||||
except urllib.error.HTTPError as exc:
|
||||
if exc.code == 403:
|
||||
return {
|
||||
"url": url,
|
||||
"ok": True,
|
||||
"status": exc.code,
|
||||
"head_status": head_status,
|
||||
"blocked": True,
|
||||
"error": "blocked automated request; verify in browser",
|
||||
}
|
||||
return {"url": url, "ok": False, "status": exc.code, "head_status": head_status, "error": str(exc)}
|
||||
except Exception as exc: # noqa: BLE001
|
||||
return {"url": url, "ok": False, "status": head_status, "error": str(exc)}
|
||||
|
||||
|
||||
def iter_urls():
|
||||
for collection in ("jobs", "papers", "industry"):
|
||||
for path in sorted((ROOT / collection / "items").glob("*.md")):
|
||||
if path.name == "README.md":
|
||||
continue
|
||||
meta = parse_frontmatter(path.read_text(encoding="utf-8"))
|
||||
for key in URL_KEYS:
|
||||
value = meta.get(key)
|
||||
if value:
|
||||
yield path.relative_to(ROOT).as_posix(), key, str(value)
|
||||
|
||||
|
||||
def main() -> int:
|
||||
parser = argparse.ArgumentParser()
|
||||
parser.add_argument("--json", action="store_true")
|
||||
parser.add_argument("--timeout", type=int, default=12)
|
||||
args = parser.parse_args()
|
||||
|
||||
results = []
|
||||
for path, key, url in iter_urls():
|
||||
result = check_url(url, args.timeout)
|
||||
result.update({"path": path, "key": key})
|
||||
results.append(result)
|
||||
|
||||
if args.json:
|
||||
print(json.dumps(results, ensure_ascii=False, indent=2))
|
||||
else:
|
||||
for result in results:
|
||||
marker = "BLOCKED" if result.get("blocked") else ("OK" if result["ok"] else "FAIL")
|
||||
print(f"{marker} {result.get('status')} {result['path']} {result['key']} {result['url']}")
|
||||
return 1 if any(not result["ok"] for result in results) else 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
raise SystemExit(main())
|
||||
Executable
+242
@@ -0,0 +1,242 @@
|
||||
#!/usr/bin/env python3
|
||||
"""Create a new collection item draft."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import re
|
||||
from datetime import date
|
||||
from pathlib import Path
|
||||
|
||||
|
||||
ROOT = Path(__file__).resolve().parents[2]
|
||||
|
||||
|
||||
TEMPLATES = {
|
||||
"jobs": """# Job: {title}
|
||||
|
||||
---
|
||||
type: job
|
||||
company:
|
||||
role: {title}
|
||||
location:
|
||||
source_url: {url}
|
||||
source_name:
|
||||
source_quality: unknown
|
||||
collected_at: {today}
|
||||
posted_at:
|
||||
first_seen_at: {today}
|
||||
last_checked_at: {today}
|
||||
snapshot_type: new
|
||||
job_code:
|
||||
previous_snapshot:
|
||||
status: unknown
|
||||
level:
|
||||
team:
|
||||
business_area:
|
||||
employment_type:
|
||||
salary:
|
||||
skills:
|
||||
-
|
||||
topics:
|
||||
-
|
||||
models:
|
||||
-
|
||||
related_papers:
|
||||
-
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
relevance: medium
|
||||
---
|
||||
|
||||
## Source Snapshot
|
||||
|
||||
- source:
|
||||
- collected_at: {today}
|
||||
- availability:
|
||||
|
||||
## Change Snapshot
|
||||
|
||||
- first_seen_at: {today}
|
||||
- last_checked_at: {today}
|
||||
- snapshot_type: new
|
||||
- previous_snapshot:
|
||||
- changed_fields:
|
||||
|
||||
## Raw JD Summary
|
||||
|
||||
-
|
||||
|
||||
## Extracted Signals
|
||||
|
||||
- hard requirements:
|
||||
- preferred skills:
|
||||
- hidden signals:
|
||||
- business direction:
|
||||
|
||||
## Knowledge Links
|
||||
|
||||
- concepts:
|
||||
- papers:
|
||||
- experiments:
|
||||
- projects:
|
||||
|
||||
## Gaps for Us
|
||||
|
||||
-
|
||||
""",
|
||||
"papers": """# Paper: {title}
|
||||
|
||||
---
|
||||
type: paper
|
||||
title: {title}
|
||||
authors:
|
||||
year:
|
||||
venue:
|
||||
url: {url}
|
||||
code_url:
|
||||
source:
|
||||
collected_at: {today}
|
||||
status: queued
|
||||
relevance: medium
|
||||
topics:
|
||||
-
|
||||
methods:
|
||||
-
|
||||
benchmarks:
|
||||
-
|
||||
models:
|
||||
-
|
||||
datasets:
|
||||
-
|
||||
related_concepts:
|
||||
-
|
||||
related_jobs:
|
||||
-
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
---
|
||||
|
||||
## One-line Takeaway
|
||||
|
||||
-
|
||||
|
||||
## Problem
|
||||
|
||||
-
|
||||
|
||||
## Core Idea
|
||||
|
||||
-
|
||||
|
||||
## Evidence
|
||||
|
||||
-
|
||||
|
||||
## Useful For Us
|
||||
|
||||
-
|
||||
|
||||
## Follow-up Experiments
|
||||
|
||||
-
|
||||
""",
|
||||
"industry": """# Industry: {title}
|
||||
|
||||
---
|
||||
type: industry
|
||||
company:
|
||||
team:
|
||||
title: {title}
|
||||
url: {url}
|
||||
source_name:
|
||||
source_type: other
|
||||
source_quality: unknown
|
||||
published_at:
|
||||
collected_at: {today}
|
||||
status: queued
|
||||
topics:
|
||||
-
|
||||
implementation_signals:
|
||||
-
|
||||
product_area:
|
||||
-
|
||||
models:
|
||||
-
|
||||
tools:
|
||||
-
|
||||
benchmarks:
|
||||
-
|
||||
related_papers:
|
||||
-
|
||||
related_jobs:
|
||||
-
|
||||
related_experiments:
|
||||
-
|
||||
related_projects:
|
||||
-
|
||||
evidence_level: medium
|
||||
relevance: medium
|
||||
---
|
||||
|
||||
## One-line Takeaway
|
||||
|
||||
-
|
||||
|
||||
## What They Built or Claimed
|
||||
|
||||
-
|
||||
|
||||
## Technical Signals
|
||||
|
||||
- architecture:
|
||||
- tool use:
|
||||
- memory:
|
||||
- evaluation:
|
||||
- safety:
|
||||
- deployment:
|
||||
- data:
|
||||
|
||||
## Evidence Quality
|
||||
|
||||
-
|
||||
""",
|
||||
}
|
||||
|
||||
|
||||
def slugify(text: str) -> str:
|
||||
text = text.lower().strip()
|
||||
text = re.sub(r"[^a-z0-9\u4e00-\u9fff]+", "-", text)
|
||||
text = re.sub(r"-+", "-", text).strip("-")
|
||||
return text or "untitled"
|
||||
|
||||
|
||||
def main() -> int:
|
||||
parser = argparse.ArgumentParser()
|
||||
parser.add_argument("collection", choices=sorted(TEMPLATES))
|
||||
parser.add_argument("slug")
|
||||
parser.add_argument("--title", default=None)
|
||||
parser.add_argument("--url", default="")
|
||||
parser.add_argument("--date", default=date.today().isoformat())
|
||||
args = parser.parse_args()
|
||||
|
||||
title = args.title or args.slug.replace("-", " ").title()
|
||||
filename = f"{args.date}-{slugify(args.slug)}.md"
|
||||
path = ROOT / args.collection / "items" / filename
|
||||
if path.exists():
|
||||
raise SystemExit(f"refusing to overwrite existing file: {path}")
|
||||
path.parent.mkdir(parents=True, exist_ok=True)
|
||||
path.write_text(
|
||||
TEMPLATES[args.collection].format(title=title, url=args.url, today=args.date),
|
||||
encoding="utf-8",
|
||||
)
|
||||
print(path.relative_to(ROOT))
|
||||
return 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
raise SystemExit(main())
|
||||
Reference in New Issue
Block a user