Clarify text detectors and reproducibility in dataset card

README.md

@@ -98,8 +98,9 @@ Each row in `data/*.jsonl` is one episode:
 - **Model input:** `context` and `turns`. Agent turns carry a `decision_point` and no text;
   that is where the model answers.
 - **Answer key:** `family`, `difficulty` and `decisions`. **Never put these in the prompt.**
-- **Customer follow-ups are scripted.** They do not react to the model, so rollouts are
-  deterministic and evaluation is reproducible.
+- **Customer follow-ups are scripted.** They do not react to the model, so the customer
+  path is fixed. Model serving can still vary at temperature 0; the published results can
+  be reproduced offline by re-scoring the saved outputs in [support-rl](https://github.com/karti-ai/support-rl).
 
 `raw/` holds the original two-file layout (`*.tickets.jsonl` = model input,
 `*.targets.jsonl` = answer key, joined by `id`). `policies/policies.md` is the policy pack,
@@ -123,7 +124,7 @@ The reward is computed in three stages, in this order:
 Over-escalation is penalised as hard as under-escalation. Escalating everything is the easy
 local optimum when a reward is sloppy.
 
-The full design, including the honest limits of the two text detectors, is in
+The full design, including the honest limits of the three text detectors, is in
 `docs/EVALUATOR.md`. The reference scorer is open source in
 [support-rl](https://github.com/karti-ai/support-rl); `docs/EVALUATOR.md` is its specification. Three worked episodes set the
 quality bar in `docs/GOLD-EXAMPLES.md`.
@@ -177,8 +178,8 @@ to match it. Every change is listed in `docs/REVIEW-LOG.md`.
   amendments that the SEC approved in April 2026 (Regulatory Notice 26-10, effective
   2026-06-04). The pack is fictional house policy. Do not treat it as current regulation.
 - **English only, text only, one fictional firm.**
-- **Two detectors are heuristic.** Investment advice and unstated timelines are caught by
-  text detectors with known limits (see `docs/EVALUATOR.md`). The advice detector misses
+- **Three checks use text detectors.** Disclosure before verification, investment advice,
+  and unstated timelines are caught by text detectors with known limits (see `docs/EVALUATOR.md`). The advice detector misses
   indirect recommendations ("it's probably wise to…"). The timeline detector cannot tell
   CMP-04's acknowledgement window from an invented resolution time of the same length.