# Council deliberation for round 001

Evidence status: illustrative. The three members represent distinct multimodal models. Their statements and the orchestrator synthesis are authored examples.

The input packet contains the four snapshot images, P001, S001, generated metadata, F01–F04, and the scoped decisions H01–H02. The full Discovery references are outside this packet.

## Independent proposals

| Member | Proposal |
|---|---|
| A | **Amend R02.** F01 shows that D1's vertical-axis text was promoted to the title despite the separate caption above the plot. Distinguishing the two roles would prevent that substitution. Use a short generic title when a data snapshot has no title to keep the field populated. Check whether C1 supports that fallback. |
| B | **Amend R03.** F02 shows that Coef and SE are column headings under both lag specifications, yet the output makes SE a standalone variable. Attach those statistics to the represented variables and preserve the lag groupings. Check whether a less explicit table layout would support the same association. |
| C | **Add a unit rule.** F04 shows that C2's bars have numerical labels but the axis lacks a unit label. Distinguish scale and printed values from the unit and entity measured. Leaving the unit unspecified whenever the axis lacks a unit would avoid the inference in C2. Check whether units are stated elsewhere in the data snapshot. |

## Cross-critiques

| Member | Critiques of the other proposals |
|---|---|
| A | B should constrain the attachment rule to cases where the layout supports it. C should allow a unit clearly stated in a caption or row heading. |
| B | A's generic title would supply wording C1 does not show. C's proposed restriction is too broad for mixed-unit tables. |
| C | A should treat the document label separately. B should preserve the visible cross-reference in D2 and avoid expanding what the stars mean, consistent with H01. |

## Revised proposals

- A retains the amendment to R02 and drops generic title substitution after B's C1 counterexample. A distinguishes the caption from axes and records a printed document label separately, following C's critique.
- B retains the amendment to R03 and makes the visual association a condition. B retains H01 as review guidance for D2.
- C limits the new rule to supported unit-to-concept associations. C2's integer scale alone cannot establish the counted entity.

## Orchestrator synthesis

Accept the scoped amendments to R02 and R03 and add R05. Keep the existing evidence and transcription rules R01 and R04. Reject the generic title fallback and the blanket ban on units when the axis lacks a label. Record the rationale and open validation question in the findings and S002. Emit P002 as the next policy candidate. Leave P001 and its extraction results available for audit.
