Model Releases
Working Notes on Late Interaction Dynamics: Analyzing Targeted Behaviors of Late Interaction Models
arXiv:2603.26259v2 Announce Type: replace-cross Abstract: While Late Interaction models exhibit strong retrieval performance, many of their underlying dynamics remain understudied, potentially hiding
arXiv:2603.26259v2 Announce Type: replace-cross Abstract: While Late Interaction models exhibit strong retrieval performance, many of their underlying dynamics remain understudied, potentially hiding performance bottlenecks. In this work, we focus on two topics in Late Interaction retrieval: a length bias that arises when using multi-vector scoring, and the similarity distribution beyond the best scores pooled by the MaxSim operator. We analyze these behaviors for state-of-the-art models on the NanoBEIR benchmark. Results show that while the theoretical length bias of causal Late Interaction models holds in practice, bi-directional models can also suffer from it in extreme cases. We also note that no significant similarity trend lies beyond the top-1 document token, validating that the MaxSim operator efficiently exploits the token-level similarity scores.
Related
- ProRank: Prompt Warmup via Reinforcement Learning for Small Language Models Reranking
- Document-as-Image Representations Fall Short for Scientific Retrieval
- WikiSeeker: Rethinking the Role of Vision-Language Models in Knowledge-Based Visual Question Answering
- LaMI: Augmenting Large Language Models via Late Multi-Image Fusion
Source: arXiv cs.CL | 2026-04-16