Tag
openjev is a model based on Qwen3.5 trained for entailment tasks, enabling applications in reranking, grading, and real-time game playing, with the v2 version adding multi-modal capabilities and improved zero-shot performance.
LAMAR is a language-aware multilingual cross-encoder reranker that uses English-anchored relevance distillation and preference alignment to prioritize documents in the same language as the query while preserving semantic relevance, achieving strong performance on multilingual benchmarks.
CityBehavEx is a scalable LLM-assisted urban simulation platform that combines established human mobility models with fine-tuned cross-encoders to generate realistic, empirically validated mobility patterns for city-sized populations, demonstrating 100,000 agents over 75 days in under one hour on a single consumer GPU.
Archex is a new open-source code RAG tool that improves retrieval by combining hybrid search (BM25F + dense embeddings), cross-encoder reranking, and dependency-graph expansion, achieving much higher recall and token efficiency than pure embeddings-based approaches.
ConvMemory v2 is a recall-preserving reranker that reorders the top-10 candidates from ConvMemory v1 using a fine-tuned cross-encoder, improving MRR on the LoCoMo benchmark while preserving recall.
A practical fix for RAG hallucination caused by noisy retrieval: use cross-encoder re-ranking to filter chunks with a score > 1.5, improving relevance from -0.28 to +3.80 on average.
Introducing the Ettin Reranker family: six new state-of-the-art CrossEncoder rerankers at various sizes, built on ModernBERT encoders, with open-source data and training recipe.