Update README.md

README.md

@@ -135,7 +135,7 @@ Three stages, all reproducible from
    extraction with a 9-rule system prompt + 3 in-context examples
    covering null window, vulnerable-with-time-stripped, and
    ephemeral-filtered cases.
-3. **Quality judge (local Qwen2.5-27B + JSON-schema constraint):** per-
+3. **Quality judge (local Qwen3.5-27B + JSON-schema constraint):** per-
    extraction verdict on an 11-mode failure taxonomy. Only extractions
    passing all 11 modes survive into this dataset.
 
@@ -185,21 +185,6 @@ for qualitative inspection.
   modes (invented_from_null, ephemeral_detail). Some chosens with
   subtle issues may have slipped through.
 
-## Suggested training
-
-Qwen2.5-Instruct base, LoRA via Unsloth or TRL `SFTTrainer`. The
-messages format is directly consumable by Unsloth's
-`apply_chat_template` + `train_on_responses_only`.
-
-Reasonable hyperparameters as a starting point:
-- 2-3 epochs
-- LR 2e-4, cosine, 5% warmup
-- LoRA rank 16-32, alpha = rank, target all linear layers
-- Batch size 4 (per device) × grad-accum 4 = effective 16
-
-JSON-schema-constrained inference is recommended at deploy time
-(llama.cpp `json_schema` field on `/completion`).
-
 ## Contact
 
 [Schneewolf Labs LLC](mailto:contact@schneewolflabs.com) — flammen.ai