AI Science Tracker
EN

Heads That Write, Not Just Point: Image Retrieval Heads in Vision-Language Models

Affiliation not stated in the reviewed material · 3 Nov 2026 · computer science

Discovery

Heads That Write, Not Just Point: Image Retrieval Heads in Vision-Language Models

Source check: Source checked. Not independent scientific confirmation.

Finding: Head-level probes and targeted ablations identify a specific pointer-writer intersection associated with multi-image inference behavior.

Technology: Attention mass identifies pointer heads

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Attention mass identifies pointer heads
Test and result
Attention mass identifies pointer heads; a trained readout identifies writer heads; targeted ablations test causal importance.; Head-level probes and targeted ablations identify a specific pointer-writer intersection associated with multi-image inference behavior.; Targeted head ablations compare pointer-writer overlap, single-criterion heads, and random controls across haystack sizes.
Limitation / still unknown
External generalization, independent replication and downstream real-world impact were not established by the reviewed source.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/html/2611.04903v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 3/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Open in the full catalog for filters, charts and related discoveries. Browse computer science discoveries.

Related catalog entries

Links use documented relationships where available, plus shared catalog context. A shared topic or model family does not establish a scientific dependency.