#NLI
2 related articles
·3 min
Hybrid Retrieval + NLI + Open-Source LLMs for Scientific Claim Verification: Why Does Fusion by Rules Hurt Performance?
An experimental study constructs a 2,035-claim benchmark to compare hybrid retrieval, NLI baselines, and three open-source LLMs (Llama 3.1, Qwen3, Mistral). Qwen3 leads with Macro-F1 0.7937, while conservative NLI+LLM fusion rules degrade performance—error analysis shows disagreements often overturn correct LLM predictions.
Read more →
·2 min
Inferring L1 Backgrounds from L2 English via LLM Token Probabilities: 70B Model Achieves SOTA
New study infers L1 backgrounds from L2 English using LLM token probabilities; Llama 3.1 8B and 70B evaluated on three datasets, with 70B achieving SOTA.
Read more →