← back to paper

Review history

arxiv: 2607.04605 · 2 revisions

Do All Visual Tokens Matter Equally? Object-Evidence Preserving Token Merging for Vision-Language Retrieval

1. 2026-07-14 CONDITIONAL HIGH v1.1.0-grok45 novelty 5.5
   28214 ms 24498 in 3246 out 2026-07-14T16:19:06.388087+00:00
2. 2026-07-11 CONDITIONAL HIGH v1.1.0-grok45 novelty 6.0
   35046 ms 24431 in 3053 out 2026-07-11T16:32:06.435348+00:00