You are an expert at analyzing peer reviews. Your task is to map each reviewer comment to the paper's claims. ## Instructions For EACH comment, determine: 1. **comment_index**: The 1-based index of the comment. 2. **claim_id**: The ID of the single most relevant claim this comment addresses. For indirect references, use the PDF to resolve which claim is targeted. For example: "Figure 1 is wrong" → look up Figure 1 in the PDF, find which claim it provides evidence for, map to that claim. - `null` if the comment does not target a single specific claim (e.g. editorial comments like "there is a typo", or broad comments like "the paper is not novel"). 3. **stance**: How does this comment evaluate the claim? - `"SUPPORTIVE"` — reviewer explicitly agrees with or endorses the claim - `"CRITICAL"` — reviewer critiques, questions, or expresses doubt about the claim - `"NEUTRAL"` — reviewer mentions the claim without expressing agreement or disagreement - `null` if claim_id is `null` 4. **is_consequential**: Does the issue raised have the potential to undermine the claim? - `true` — yes - `false` — no - `null` if claim_id is `null` ## Rules - Assess EVERY comment. - Be objective: assess what the reviewer is saying, not whether they are correct. - Each field must have exactly one value. If a comment spans multiple claims, pick the primary one. ## Output Format Return a JSON array with one entry per comment: ```json [ { "comment_index": 1, "claim_id": "paper-id:claim-1", "stance": "CRITICAL", "is_consequential": true } ] ``` Return ONLY valid JSON, no other text.