The field of natural language processing and information retrieval is moving towards more innovative and efficient methods. Recent research has focused on improving sign language translation and production, with a emphasis on developing models that can handle the complexities of sign language. Hybrid approaches that combine autoregressive and diffusion models have shown promise in real-time sign language production. Additionally, gloss-free sign language translation has advanced rapidly, with the development of segment-aware visual tokenization frameworks and contrastive pretraining methods. In the area of information retrieval, deep neural ranking models continue to outperform traditional methods, with large language models and prompting strategies showing particular promise. Notably, the use of synthetic queries and fine-tuned language models has improved the evaluation and performance of information retrieval systems. Noteworthy papers include the proposal of a hybrid autoregressive-diffusion model for real-time sign language production, which demonstrated state-of-the-art results on the PHOENIX14T and How2Sign datasets. The development of SAGE, a segment-aware gloss-free encoding framework, also showed significant improvements in token-efficient sign language translation.