A comprehensive index of artificial intelligence and machine-learning research with AI-generated summaries, citation metrics, and direct links to papers and code.
Orion Weller, Michael Boratko, Iftekhar Naim, et al.
This paper proves that single-vector embedding models have a fundamental dimension-dependent limit on the number of distinct top-k subsets they can represent, and shows this limitation arises even in simple realistic queries.
B. Chen, Gregory W. Wornell
Introduces quantization index modulation (QIM) and distortion-compensated QIM for provably good digital watermarking with optimal rate-distortion-robustness tradeoffs.
Iulian Vlad Serban, Alessandro Sordoni, Yoshua Bengio, et al.
Extends hierarchical recurrent encoder-decoder networks to open-domain dialogue, showing competitiveness with n-gram models and improvement via bootstrapping from QA pairs and pretrained embeddings.
Unknown
This paper proposes Structural LM, which extends BERT by incorporating 1D and 2D cell-level embeddings for structured text understanding.
Unknown
LAMBERT enhances RoBERTa with layout embeddings and relative bias for layout-aware document understanding without raw images.
Xinkun Huang, Jinyan Sang, Changrong Jin, et al.
Extends BERT with 2-D position and image embeddings for layout-aware document understanding.
Daniel Bolya, Po-Yao Huang, Peize Sun, et al.
Perception Encoder is a vision encoder trained via contrastive vision-language learning that achieves state-of-the-art results across diverse tasks by extracting strong general embeddings from intermediate layers.
Unknown
Leverages Gemini LLM to generate generalizable multilingual text and code embeddings via synthetic data and filtering.
Unknown
mmE5 is a multimodal multilingual embedding model trained on synthetic data via a framework that ensures broad task/language coverage, robust cross-modal alignment, and high fidelity through self-evaluation and refinement.
Unknown
vdr Embeddings improve visual document retrieval via synthetic data training and Matryoshka Representation Learning.
Saba Sturua, Isabelle Mohr, Mohammad Kalim Akram, et al.
Jina Embeddings v3 is a 570M parameter multilingual embedding model supporting 8192-token contexts, task-specific LoRA adapters, and flexible dimension reduction via Matryoshka learning.
Unknown
A framework for customizing LLM embeddings to enable significant dimensionality reduction without performance loss.