Tag
The paper reveals that cross-lingual harmfulness representations in large language models degrade with resource tier when evaluated using hard negatives, challenging previous conclusions that safety representations survive translation.
This paper introduces UniME-R1, an embedder-adviser framework for unified multimodal retrieval that generates Retrieval-Centric Chain-of-Thought (RC-CoT) conditioned on retrieval feedback, improving retrieval performance by learning from hard negatives.
This paper investigates principles of concept representation in sentence encoders through the lens of compositional semantics, identifying four key principles: fine-tuning recalibrates latent geometry, semantic signal concentrates in the final layer, hard negatives improve discrimination but not ranking, and supervision effectiveness depends on composition type.