A new research paper introduces Wontopos Tablet 2, a long-term memory engine for language models designed for multilingual and multimodal retrieval without relying on lexical matching. The system demonstrated strong performance on text benchmarks like LongMemEval-S, achieving 95.7% accuracy, and on the BEAM-1M dataset with 67.5% accuracy. In multimodal tests, Wontopos Tablet 2 significantly outperformed BM25 in cross-lingual retrieval of photographs, achieving 95.2% recall compared to BM25's 19.0%, and showed better language independence on captionless images. AI
IMPACT Advances memory retrieval capabilities for LLMs, potentially improving performance in multilingual and multimodal applications.
RANK_REASON The cluster contains a research paper detailing a new system and its performance on benchmarks.
Read on arXiv cs.IR (Information Retrieval) →
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →