# Paper Insights

status: seed

这个文件沉淀论文阅读后的研究判断。单篇论文放在 `papers/items/`，这里记录跨论文的趋势和结论。

## Current Snapshot

- 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 |
| --- | --- | --- | --- |
| 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: 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