Downloads · 30 days
4
2% of all-time downloads
cs-552-2026-Clanker-Scientists
safeguard-deberta-ragtruth-v1
cs-552-2026-Clanker-Scientists/safeguard-deberta-ragtruth-v1
safeguard-deberta-ragtruth-v1 is a text classification model from cs-552-2026-Clanker-Scientists. Use it when you need a label for a piece of text. The card lists the license as mit.
DeBERTa-v3-large fine-tuned on RAGTruth for generated-answer faithfulness. 2-class head: 0 = supported, 1 = hallucinated. Used as the pipeline's grounding verifier: premise = cited context, hypothesis = drafted answer…
All-time downloads
265
Public
Parameters
435M
10.4 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors1.7 GB · 99%
At a glance
- Task
- Text Classification
- License
- mit
- Model type
- deberta-v2
- Access
- Public
- Created
- Jun 4, 2026
- Updated
- Jun 8, 2026
- SHA
- da89cd92
Try a prompt
Base models
Model card
From the Hugging Face model README
Safeguard — DeBERTa-v3-large, RAGTruth fine-tuned (faithfulness verifier)
DeBERTa-v3-large fine-tuned on RAGTruth for generated-answer faithfulness.
2-class head: 0 = supported, 1 = hallucinated. Used as the pipeline's grounding
verifier: premise = cited context, hypothesis = drafted answer, argmax == 0 -> supported.
M3 results
| dataset | macro-F1 |
|---|---|
| RAGTruth (in-domain) | 0.80 (hallucination P/R 0.71 / 0.79) |
| CogniBench (transfer) | 0.66 |
| MEMERAG-FR (transfer) | 0.58 |
The earlier ContractNLI proxy checkpoint scored 0.838 on its proxy but only 0.52 zero-shot on real RAGTruth faithfulness — use this checkpoint for grounding generated answers.
from transformers import AutoModelForSequenceClassification, AutoTokenizer
tok = AutoTokenizer.from_pretrained("cs-552-2026-Clanker-Scientists/safeguard-deberta-ragtruth-v1")
mdl = AutoModelForSequenceClassification.from_pretrained("cs-552-2026-Clanker-Scientists/safeguard-deberta-ragtruth-v1")
# supported iff mdl(**tok(context, answer, return_tensors="pt")).logits.argmax(-1) == 0
- Task
- Text Classification
- Type
- deberta-v2
- License
- mit
- Created
- Jun 4, 2026
- Updated
- Jun 8, 2026