Token pooling-aware training for multi-vector embedding models improves retrieval quality by teaching the encoder to preserve information when its vectors are merged, helping to retrieve better from the same storage budget. By applying a loss wrapper to a multi-vector retrieval loss, pooling-aware training can be used with existing models, resulting in a 97.8% retention of unpooled NanoBEIR nDCG@10 with a 4x reduction in document vectors. AI summary
Firehose
Filtered to tagged “token pooling” · clear filters
Browse: People · Companies · Papers · Podcasts · Hacker News · Deep dives
Browse by tag
artificial intelligence 63continual learning 22AI 19agentic coding 17open-weight models 14reinforcement learning 12AI agents 11AI safety 8datasette 7machine learning 7natural language processing 7benchmarks 6language models 6Nvidia 6Reinforcement learning 6robotics 6deep learning 5Diffusion models 5large language models 5open-source 5Agentic AI 4AI ethics 4AI Gateway 4Databricks 4decision models 4Gemini 4general 4Jensen Huang 4Recursive self-improvement 4software development 4