card: fine-tuned model repos

README.md

@@ -68,6 +68,21 @@ Consensus columns: `vote_all__pred`, `vote7__pred`, `agree4__pred`, `agree2__pre
 
 Measured on the held-out gold test set: the best single deployed signal (Gemma-4-31B-it) reaches F1 0.844 (0.944 on predicates seen during development, 0.776 on unseen ones). The published `agree3` label covers 66.8% of the test rows and on them reaches precision 0.899, recall 0.934, F1 0.916 (95% CI 0.886 to 0.944). Use the per-signal scores to choose a different precision/recall operating point.
 
+## Fine-tuned models
+
+The HebNLI-fine-tuned denoisers behind the released signals and the per-strategy winners are on the Hub (private until publication):
+
+| Model | Role in HeRE | Repo |
+|---|---|---|
+| DictaLM-3.0-24B (Thinking) HebNLI LoRA | Strategy 2; one of the three `agree3` models behind the published label | https://huggingface.co/ronke21/hebnli-dictalm-3.0-24b-lora |
+| NeoDictaBERT HebNLI | Strategy 1; deployed signal `encoder_nli__neodictabert` | https://huggingface.co/ronke21/hebnli-neodictabert |
+| XLM-RoBERTa-large HebNLI | Strategy 1; deployed signal `encoder_nli__xlmroberta` | https://huggingface.co/ronke21/hebnli-xlm-roberta-large |
+| XLM-RoBERTa-XL HebNLI | Strategy 1 best (F1 0.779) | https://huggingface.co/ronke21/hebnli-xlm-roberta-xl |
+| Qwen3.5-35B-A3B-Base HebNLI LoRA | Strategy 2 best (F1 0.873) | https://huggingface.co/ronke21/hebnli-qwen3.5-35b-a3b-lora |
+| Gemma-4-31B HebNLI LoRA | Strategy 2 (F1 0.849) | https://huggingface.co/ronke21/hebnli-gemma-4-31b-lora |
+
+The Strategy 4 signals (Gemma-4-31B-it, Gemma-3-27B-it, DictaLM-3.0-24B-Base) are prompted public models; prompts are in the code repository.
+
 ## Contamination note
 
 The 500 test rows were drawn from the silver corpus; the 499 that match a silver row are **withheld from every silver and KG file** here. Do not add them back when training.