DeBERTa-v3-base-mnli-fever-anli
A DeBERTa-v3 discriminative natural language inference model fine-tuned on MNLI, FEVER and ANLI. It is effective at entailment judgments over single-sentence evidence but struggles to integrate evidence across multiple sentences.
Timeline (last 90 days)
On the scientific claim verification benchmark, Qwen3 8B achieves the highest Macro-F1 (0.7937), Llama 3.1 8B reaches 0.7204, Mistral 7B reaches 0.6388, and the NLI baseline (DeBERTa-v3-base-mnli-fever-anli) reaches only 0.3157.
DeBERTa-v3-base-mnli-fever-anli is a discriminative NLI model fine-tuned on MNLI, FEVER and ANLI, and is essentially a sentence-pair-level classifier.
All Facts (2)
On the scientific claim verification benchmark, Qwen3 8B achieves the highest Macro-F1 (0.7937), Llama 3.1 8B reaches 0.7204, Mistral 7B reaches 0.6388, and the NLI baseline (DeBERTa-v3-base-mnli-fever-anli) reaches only 0.3157.
50%UnverifiedDeBERTa-v3-base-mnli-fever-anli is a discriminative NLI model fine-tuned on MNLI, FEVER and ANLI, and is essentially a sentence-pair-level classifier.
50%