Dual-View Training for Instruction-Following Information Retrieval
2026-04-22T08:52:17Z•92322003fee84e15e897efd3bdf9f5cb7fe9258444818490b31fcd8898636b06
CS3CTR scalingColBERT diagnosticsFeatGEOFollowIRGraphRAG-IRLLoopCTRSTK-Adaptercounterfactual explanationsdata synthesisdiagnosable retrievalgenerative citation optimizationinformation retrievalinstruction-following retrievalinverse reinforcement learningmixture-of-expertspolarity reversalrecommendation systemsreproducibilitytemporal knowledge graphstwo-tower retrievers
What happened
This RSS batch contains multiple new arXiv submissions (Apr 22, 2026) in information retrieval, recommender systems, and related ML methods. Highlights include: (1) Dual-View Training for instruction-following IR — a polarity-reversal data synthesis method that improves instruction sensitivity of retrievers (reported +45% on FollowIR); (2) STK-Adapter — an adapter integrating evolving temporal KG encoders and LLMs with MoE experts for TKG extrapolation; (3) FeatGEO — a feature-level multi-objective optimization framework to improve citation visibility for generative answer engines; (4) GraphRÅ
Why it matters
A reviewed impact interpretation has not been published for this record.
Evidence and limitations
- Source ID
- arxiv_cs_ir
- Record identifier
- 92322003fee84e15e897efd3bdf9f5cb7fe9258444818490b31fcd8898636b06
- Enrichment time
- 2026-04-22T08:52:17Z
- AI-assisted enrichment
- Yes
This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.