ResearchBenchmarksVeracity Inference on PRONTOQA 3-hop (test)Follow95.6Hamming SimilarityAVI65.12873.03980.9588.861May 17, 2025Evaluation ResultsMethodMethodLinksHamming SimilarityAVIBase LLM=Qwen 8BBase LLM=Qwen 8B2025.0595.6AVIBase LLM=Qwen 4BBase LLM=Qwen 4B2025.0588.6Many2ManyBase LLM=Qwen 4BBase LLM=Qwen 4B2025.0571Many2ManyBase LLM=Qwen 8BBase LLM=Qwen 8B2025.0566.3