ReasonIR: Training Retrievers for Reasoning Tasks

By Rulin Shao, Rui Qiao, Varsha Kishore, Niklas Muennighoff, Xi Victoria Lin, Daniela Rus, K. Low, Sewon Min, Wen-Tau Yih, Pang Wei Koh and 1 more. 2025

Subfields: Information Retrieval & Search, Prompting & Reasoning

Citations in the corpus

• BRIDGE: Multimodal-to-Text Retrieval via Reinforcement-Learned Query Alignment (2026, unknown): “Subsequent work has sought to address this gap through reasoning-aware fine-tuning [10, 36] and iterative query expansion [19, 41].”
• MARVEL: Multimodal Adaptive Reasoning-intensiVe Expand-rerank and retrievaL (2026, unknown): “Subsequent work has sought to bridge this gap through reasoning-aware fine-tuning [9, 29] and iterative query expansion [19, 33].”
• Seq vs Seq: An Open Suite of Paired Encoders and Decoders (2025, unknown): “As Ettin models have been trained with instructions during pre-training, it is likely they are also more capable for instruction-based retrieval (Shao et al., 2025; Weller et al., 2024; 2025), however, we leave that for future work.”
• A Survey of Model Architectures in Information Retrieval (2025, unknown): “The commonly used smoothing methods (Zhai and Lafferty, 2004; Zhai et al., 2008) include Jalinek-Mercer smoothing, Dirichlet smoothing, etc.”
• HIVE: Query, Hypothesize, Verify An LLM Framework for Multimodal Reasoning-Intensive Retrieval (2026, unknown): “Subsequent work has sought to address this gap through reasoning-aware fine-tuning [9, 30], iterative query expansion [17, 34], and LLM-based reranking [4, 24, 32, 37, 43].”
• MM-BRIGHT: A Multi-Task Multimodal Benchmark for Reasoning-Intensive Retrieval (2026, unknown): “To address these challenges, specialized reasoning-enhanced retrievers have been proposed: ReasonIR [23] leverages chain-of-thought data for training, while Rader [6] and DiVeR [13] incorporate reasoning traces into the retrieval process.”