Tabbit
活动资源博客模型
Tabbit LogoTabbit

Tabbit — 为你工作的 AI 浏览器

主题资源

  • AI Browser Resources
  • Agentic Browser Resources
  • Browser Downloads and Install Guides
  • Browser Comparisons
  • AI Browser Alternatives
  • Browser Productivity Resources

热门指南

  • AI Browser
  • Agentic Browser Download
  • Best AI Browser 2026: Top 9 Tested & Ranked
  • AI Browser Download
  • Free AI Browser
  • Best AI Browser 2026
  • AI Browser Comparison 2026
  • AI Browser for Windows
  • AI Browser for Mac
  • Chrome Alternative 2026

活动

  • 别装了,你在《牛来》里早有原型
  • Tabbit 妙招大赛
  • KPOP SBTI 饭圈人格测试
  • Tabbit 校园共创者计划
  • fifi 的论文文献妙招精选
  • 用户问卷

关于

  • Tabbit 博客
  • 媒体报道
简体中文
简体中文English
提示词与工作流

GPT-5.2 Chat · prompting-guide

GPT-5.2's Structured Outputs and Ambiguity Self-Check Prompt

The official guide emphasizes an output contract, ambiguity handling, and self-checks; this detail targets one parseable structured decision.

来源待核验OpenAI API; source material, JSON schema, and a conflict example.

前置条件与输入

  • Task goal and source material
  • Output format or schema
  • Acceptance rules

完整可复制模板

Structured decision prompt block

Tabbit 编辑改写;非来源原文
Source records: {{SOURCE_RECORDS}}
Output schema: {{DECISION_SCHEMA}}
Ambiguity case: {{AMBIGUITY_CASE}}
Validation rule: {{VALIDATION_RULE}}

运行前仍需替换: {{SOURCE_RECORDS}}, {{DECISION_SCHEMA}}, {{AMBIGUITY_CASE}}, {{VALIDATION_RULE}}

Prerequisites

OpenAI API; source material, JSON schema, and a conflict example.

Task steps

Fix the inputs, output contract, and tool boundary; save real returns, errors, and screenshots after each round.

Task result

Deliver the artifact for “GPT-5.2's Structured Outputs and Ambiguity Self-Check Prompt” and list what the inputs cannot confirm.

Output and acceptance

Run the actual acceptance command and check format, critical paths, and evidence records.

Failure correction

Reproduce the smallest failing case, then narrow the input or fix tool arguments; do not treat model self-report as completion evidence.

Source and boundary

The official guide emphasizes an output contract, ambiguity handling, and self-checks; this detail targets one parseable structured decision.

查看来源研究笔记

一句话结论

把输出边界、歧义处理、长上下文复核和高风险自检写成明确区块,可让 GPT-5.2 Chat 在低推理档位下更稳定地遵守格式并减少无依据断言。

适用场景

  • 适合的任务:资料抽取、客服/研究问答、长文档定位、需要固定 Markdown 或 JSON 形状的工作流。

  • 不适合的任务:需要服务端强制 schema、权限或事实验证的场景;提示词不能替代校验和安全策略。

  • 适用的模型版本:GPT-5.2 家族;gpt-5.2-chat-latest 是 ChatGPT 对齐的快照,官方文档同时将 gpt-5.2 定位于更复杂的推理任务。

  • 适用的客户端、Agent 或 API:Chat Completions 可复用提示区块;Responses API 更适合保留多轮推理上下文。

  • 推荐的推理档位和参数:Chat 快速交互优先 reasoning.effort=none(仅适用于支持该参数的接口);输出长度用 text.verbosity=low|medium|high,实际支持以当前 API 为准。

可直接使用的内容

下面是 OpenAI 指南中可直接复用、按研究/抽取场景拼接的原始提示区块;其中字段值可按任务替换。

<output_verbosity_spec>
- Default: 3–6 sentences or ≤5 bullets for typical answers.
- For simple “yes/no + short explanation” questions: ≤2 sentences.
- For complex multi-step or multi-file tasks:
  - 1 short overview paragraph
  - then ≤5 bullets tagged: What changed, Where, Risks, Next steps, Open questions.
- Provide clear and structured responses; use lists, paragraphs and tables when helpful.
- Avoid long narrative paragraphs; prefer compact bullets and short sections.
- Do not rephrase the user’s request unless it changes semantics.
</output_verbosity_spec>

<long_context_handling>
- For inputs longer than ~10k tokens:
  - First, produce a short internal outline of the key sections relevant to the request.
  - Re-state the user’s constraints explicitly before answering.
  - Anchor claims to source sections rather than speaking generically.
- If the answer depends on fine details, quote or paraphrase those details.
</long_context_handling>

<uncertainty_and_ambiguity>
- If the question is ambiguous or underspecified, call this out and:
  - Ask up to 1–3 precise clarifying questions, OR
  - Present 2–3 plausible interpretations with clearly labeled assumptions.
- When external facts may have changed recently and no tools are available, answer generally and state that details may have changed.
- Never fabricate exact figures, line numbers, or external references.
- When unsure, prefer “Based on the provided context…” instead of absolute claims.
</uncertainty_and_ambiguity>

<high_risk_self_check>
Before finalizing an answer in legal, financial, compliance, or safety-sensitive contexts:
- Re-scan for unstated assumptions.
- Re-scan for numbers or claims not grounded in context.
- Re-scan for overly strong language such as “always” or “guaranteed”.
- Qualify any problem found and state assumptions explicitly.
</high_risk_self_check>

测试/工作流步骤

  1. 将区块放在系统或开发者消息中,把用户输入作为独立任务内容。

  2. 用同一份长文档和同一组缺失字段问题,分别测试 none 与 medium(若接口支持)。

  3. 记录格式合规率、缺失字段是否为 null/未找到、引用能否回指原文,以及总输出 token。

  4. 在高风险任务中,服务端对 JSON/schema、权限、引用和数值再做独立校验。

原始证据与数据

  • 官方指南把 GPT-5.2 描述为更强的结构化推理、工具 grounding 和多模态模型,并指出它仍对输出形状、语气和详细程度敏感。

  • 官方给出的输出约束示例为普通回答 3–6 句或不超过 5 个 bullet;长上下文阈值示例约为 10k tokens。

  • 官方给出的迁移建议是先保持提示不变建立 baseline,再固定 reasoning_effort,每次只改一个提示因素并重跑 eval。

  • 上述阈值和参数是指南中的建议,不是对所有客户端的硬保证。

适用边界

  • 该文档来自 GPT-5.2 家族指南,不能据此断言每个 gpt-5.2-chat-latest 快照都具备同样的推理/Responses 行为。

  • Chat 版本官方模型页标记为 deprecated;生产系统应先核对当前可用 snapshot 和迁移建议。

  • “内部测试/客户反馈”没有公开完整数据集、样本数和方差,不能当作独立 benchmark。

  • “内部 outline”仅是行为要求,不应要求模型泄露隐藏 chain-of-thought;只要求可核验的简短结论、引用和假设。

来源摘录或观察(仅做合规短引)

官方指南的核心要求是“Never fabricate exact figures, line numbers, or external references when you are uncertain.”(原文短引,14 词以内)。

来源与日期

OpenAI Developers / GPT-5.2 Prompting Guide · 原文日期: 未公开 · 编辑日期: 2026-09-20

阅读原始来源
变量检查

仍需替换: 4

{{SOURCE_RECORDS}}{{DECISION_SCHEMA}}{{AMBIGUITY_CASE}}{{VALIDATION_RULE}}

相关提示词

GPT-5.2 Chat's Responses, Reasoning, and Tool-Calling Configuration

相关测评

GPT-5.2 家族:官方发布基准与 Chat 定位SWE-bench 榜单:GPT-5.2 编码 Agent 对照GPT-5.2 发布后 Reddit 用户的编码与对话体验

模型深度阅读

总览 · 简体中文

GPT-5.2 Chat 是什么:价格、生命周期与迁移边界

说明 GPT-5.2 Chat 的 ChatGPT 对齐 API 路线、128K 上下文、价格、ChatGPT/Codex 生命周期与迁移选择。

GPT-5.2 Chat

在 Tabbit 中使用 GPT-5.2 Chat

请在上方所列环境中运行本指南。下载不会自动传入模板,也不代表账户已开放该模型。