Researchers have developed a system called SciClaimSeekers to retrieve and rerank scientific sources for claims made on social media. This framework combines traditional methods like BM25 with advanced LLM reranking using Qwen2.5-14B-Instruct. The system achieved a significant improvement in performance, reaching 64.36% MRR@5 on an English development set, outperforming baseline methods by over 10 points. This indicates that carefully constructed pipelines leveraging large pre-trained models can be highly effective for this task. AI
IMPACT This system could improve the verification of information spread on social media by linking claims to scholarly sources.
RANK_REASON The cluster describes a scientific paper detailing a new system and its performance on a specific task. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
