EGIP: An Evidence-Grounded Interaction Protocol for Clinical Agents

1. Analyze the Request: * Role: Professional news typesetting engineer. * Task: Convert plain text news content into an "semantic" HTML fragment. * Rules: 1. Output *only* the HTML fragment itself (no Markdown fences, no explanations). 2. Paragraphs wrapped in `

`, long text divided by `

`/`

`. 3. Lists use `

• `, quotes use `

  `, code uses `

  `, important sentences use ``.
          4.  Preserve *all* original information, do not add/delete/modify content.
          5.  Do not output document-level tags like ``, ``, ``.
      *   Input Text: A short abstract about a paper proposing EGIP (Evidence-Grounded Interaction Protocol) for clinical agents.
  
  2.  Analyze the Input Text & Determine Structure:
      *   *Sentence 1:* "Interactive clinical agents acquire evidence through patient dialogue and diagnostic examinations, yet their performance also depends on how the interaction environment exposes information, available actions, and examination results." -> Introduction/Problem statement. Paragraph 1.
      *   *Sentence 2:* "When these conditions are implicit or inconsistently controlled, final answer accuracy can conflate diagnostic reasoning with the effects of the interaction protocol." -> Problem elaboration. Paragraph 2.
      *   *Sentence 3:* "We propose the \textbf{E}vidence-\textbf{G}rounded \textbf{I}nteraction \textbf{P}rotocol (EGIP), which enforces role-specific information boundaries, exposes diagnostic actions without revealing their results, retrieves examination results from stored clinical records, and organizes acquired evidence for final decision-making." -> Solution proposal. Paragraph 3. (Note: `\textbf{...}` should be converted to `...` as per rule 3).
      *   *Sentence 4:* "By separating evidence acquisition from terminal prediction, EGIP enables different decision strategies to operate on the same observations." -> Mechanism explanation. Paragraph 4.
      *   *Sentence 5:* "We evaluate EGIP across two clinical-agent frameworks, five datasets, and six language models." -> Evaluation scope. Paragraph 5.
      *   *Sentence 6:* "EGIP improves diagnostic accuracy over native baselines across all primary comparisons." -> Main finding. Paragraph 6.
      *   *Sentence 7:* "A clinician review further shows that the proportion of automatically scored correct responses jointly judged evidence-supported rises from 10.20% to 36.55%." -> Additional finding. Paragraph 7.
      *   *Sentence 8:* "These findings support the effectiveness of EGIP as a unified interaction protocol for improving clinical-agent performance through controlled evidence access and acquisition." -> Conclusion. Paragraph 8.
  
  3.  Apply Formatting Rules:
      *   Since it's a short abstract, dividing into `
  `/`
  ` might be tricky without inventing subheadings that weren't explicitly there. However, the prompt says "长文用
  /
  划分小节" (long text divided into sections with h2/h3). This is a short text, but I can structure it logically or just group them into a few `
  ` blocks. Let's look at the flow. It's one continuous thought. I will keep it as a few `
  ` paragraphs to maintain readability without forcing artificial headings that violate "不得增删改内容" (don't add/delete/modify content - adding headings is adding structure, but maybe breaking into sections is better?). Actually, breaking it into logical sections with `
  ` for "Problem", "Solution", "Evaluation", "Conclusion" might be appropriate for "新闻排版" (news typesetting), but wait, the text doesn't have explicit markers. I will just format it into cohesive paragraphs, perhaps using `
  ` for the core problem and solution if it helps semantic structure, or just standard paragraphs. Let's stick strictly to the text. I will break it into 3-4 paragraphs to make it readable, which fits the `
  ` requirement.
      *   Wait, let's re-read carefully: "每个自然段落用
  包裹，长文用
  /
  划分小节". The input is very short. I will group the sentences logically into paragraphs.
          *   Para 1: Sentences 1-2 (Problem)
          *   Para 2: Sentence 3 (Proposed Solution) -> Convert `\textbf{E}vidence-\textbf{G}rounded \textbf{I}nteraction \textbf{P}rotocol` to `Evidence-Grounded Interaction Protocol`.
          *   Para 3: Sentences 4-5 (Mechanism & Evaluation)
          *   Para 4: Sentences 6-8 (Results & Conclusion)
      *   Let's check for lists, quotes, code, strong tags.
          *   Strong tag: `\textbf{E}vidence-\textbf{G}rounded \textbf{I}nteraction \textbf{P}rotocol` -> `Evidence-Grounded Interaction Protocol`
          *   No lists, no quotes, no code in this specific text.
  
  4.  Drafting the HTML:
      ```html
      
  Interactive clinical agents acquire evidence through patient dialogue and diagnostic examinations, yet their performance also depends on how the interaction environment exposes information, available actions, and examination results.
      
  When these conditions are implicit or inconsistently controlled, final answer accuracy can conflate diagnostic reasoning with the effects of the interaction protocol.
      
  We propose the Evidence-Grounded Interaction Protocol (EGIP), which enforces role-specific information boundaries, exposes diagnostic actions without revealing their results, retrieves examination results from stored clinical records, and organizes acquired evidence for final decision-making.
      
  By separating evidence acquisition from terminal prediction, EGIP enables different decision strategies to operate on the same observations.
      
  We evaluate EGIP across two clinical-agent frameworks, five datasets, and six language models.
      
  EGIP improves diagnostic accuracy over native baselines across all primary comparisons.
      
  A clinician review further shows that the proportion of automatically scored correct responses jointly judged evidence-supported rises from 10.20% to 36.55%.
      
  These findings support the effectiveness of EGIP as a unified interaction protocol for improving clinical-agent performance through controlled evidence access and acquisition.
      ```
      Wait, is this too many paragraphs? Yes, 8 paragraphs. Can I combine some? "长文用
  /
  划分小节" implies if it's long, use h2/h3. Since it's an abstract, grouping them into 3-4 paragraphs makes sense.
      Let's group:
      Para 1: Sentences 1-2 (Context/Problem)
      Para 2: Sentence 3 (Method)
      Para 3: Sentences 4-5 (Mechanism & Scope)
      Para 4: Sentences 6-8 (Results & Conclusion)
  
      Let's refine the grouping:
      
  Interactive clinical agents acquire evidence through patient dialogue and diagnostic examinations, yet their performance also depends on how the interaction environment exposes information, available actions, and examination results. When these conditions are implicit or inconsistently controlled, final answer accuracy can conflate diagnostic reasoning with the effects of the interaction protocol.
      
  We propose the Evidence-Grounded Interaction Protocol (EGIP), which enforces role-specific information boundaries, exposes diagnostic actions without revealing their results, retrieves examination results from stored clinical records, and organizes acquired evidence for final decision-making. By separating evidence acquisition from terminal prediction, EGIP enables different decision strategies to operate on the same observations.
      
  We evaluate EGIP across two clinical-agent frameworks, five datasets, and six language models. EGIP improves diagnostic accuracy over native baselines across all primary comparisons. A clinician review further shows that the proportion of automatically scored correct responses jointly judged evidence-supported rises from 10.20% to 36.55%. These findings support the effectiveness of EGIP as a unified interaction protocol for improving clinical-agent performance through controlled evidence access and acquisition.
  
      This looks good and respects the original text while applying the formatting rules. No ``, ``, etc. Just the fragment.
  
  5.  Final Review against constraints:
      *   Only output HTML fragment? Yes.
      *   No Markdown fences? Yes.
      *   No explanations? Yes.
      *   Paragraphs in `
  `? Yes.
      *   Long text divided by `
  `/`
  `? The text is short, so grouping into `