Repository path: workshop/regimes/R20-r1-forced.md · rendered 2026-09-09

R20 — R1 displaced marking, applied to sustained prose

v0.1, pilot, minted 2026-08-06 (S121). The regime under which the lead renders a whole passage under framework/v0.1's single recommendation, R1. It exists because every previous application of R1 has been to isolated sites — a span of two or three sentences, re-rendered once the site had already been identified as a loss. Rendering a scene whole under the brief is a different act: the translator does not know in advance which sites will be scored, and the devices have to be distributed across the passage rather than concentrated at one place.

Roles

• Translator — the lead agent, as a labeled subject. No panel call, no API cost.
• Judge — never the lead (charter §5).

Human involvement

None.

Procedure

1. Source only, exactly as R04 §1. No published rendering of the passage in any target language is read before the translation and its log are frozen.
2. The brief, quoted from framework/v0.1 §2 and not paraphrased:

Where the source marks a relation or attitude by a grammatical form the target lacks, the absence of a same-category counterpart is not the absence of the marking. Do not record the loss on the ground that the category is missing. Render the site again under a brief that requires the marking to appear, and let it fall wherever the target does mark such things — on a pronoun, a verb, an address noun, a courtesy formula, an adjective, the argument structure of the clause.

3. The marking must appear at every site where the source marks, not at a chosen few. Where the translator judges that no device is available, the site is recorded in the log as a declared failure of R1 rather than passed over.
4. The forced rendering must not merely reproduce the close one. Note (bhb): the lead matches itself at up to 37 contiguous tokens across sessions, so a R20 span that is byte-similar to its R04 counterpart has marked nothing. Each site's device is named in the log.
5. No device may assert a fact the source does not. RS-20260805h §3: the Polish suffix -ek modulates a relation, the English periphrasis a minor little clerk predicates size. Where the only available device would predicate, the site is a declared failure under step 3.
6. The log is frozen before any evaluation of the translation is designed (charter §8, A4), and names, per site, the source's device, the English device chosen, and what was given up.

What R20 is not

• Not a claim that the marked rendering is better. framework/v0.1 §2: R1 tells a translator the option exists and cannot tell them to take it, because that judgment is evidence class X3 and X3 is inadmissible while Tier D is NOT PASSED.
• Not a fluency regime. Whatever the marking costs in naturalness is R07's question, not this one.

Changelog

• v0.1 (2026-08-06, S121). Minted for E-20260806-published-loss, the first application of R1 to whole scenes. pilot until a second work is rendered under it.
• v0.1 note (2026-08-06, S122). A second application, T-kohlhaas-R20-v2, ran the brief with a declared restriction — one device category only, the archaic pronoun — in order to isolate what that category does. It is recorded there as a departure from step 3, not as a new regime, and it produced two things this page should carry: (i) the categories are alternatives, not ingredients — carrying the marking on the pronoun made the address nouns redundant rather than additional, so a rendering that applies two categories marks the same relation twice; and (ii) step 5 does not reach register. It forbids a device that asserts a fact the source does not; the archaic pronoun asserts no fact and does assert a period, dragging 15 archaic verb forms in with 41 pronoun tokens where the German is ordinary contemporary prose. Step 5's wording is left unchanged — a regime is not amended off one scene — and the gap is named here so the next application looks for it. The regime stays pilot.