AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

When Sinks Help or Hurt: Unified Framework for Attention Sink in MLLMs

Jiho Choi, Jaemin Kim, JINHWI PARK, Seunghoon Hong, Sanghwan Kim

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Attention-based visual token pruning

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
learn per-layer V-sink vs rest key scaling gate (LSG) at layer 10 with NTP loss, frozen backboneLLaVA-1.5-7B · CVBenchaccuracy delta (%p): +2.14Improved
stack five independently trained LSG gates {L3,L6,L7,L10,L19} (ablation)LSG · MMStar / CVBenchaccuracy delta (%p): +0.72 -> +1.55 MMStar; +2.14 -> +3.08 CVBenchImproved
add L19 gate on top of three stacked gates (ablation)LSG stack · MMStaraccuracy delta (%p): +1.17 -> +0.74Hurt
apply LSG at early layer L2 (ablation)LSG · MMStar / CVBench sub-tasksnon-negative sub-tasks of 10: 3 of 10 non-negative vs 8 of 10 at L10Hurt
replace prior sink methods (VAR, Sink-to-the-front, DIYSink) with LSG L10LLaVA-1.5-7B · MMStar / CVBenchaccuracy delta (%p): +0.72/+2.14 vs best prior +0.25/+0.75Improved
rebalance sink vs ordinary attention with oracle gate on fine-grained perception (ablation)LLaVA-1.5-7B · MMStar fine-grainedaccuracy gain (%p): up to +6.49Improved
rebalance sink vs ordinary attention with oracle gate on coarse perception (ablation)LLaVA-1.5-7B · MMStar coarseaccuracy gain (%p): up to -4.27Hurt
strengthen sinks in logical reasoning, early vs late layers (ablation)LLaVA-1.5-7B · MMStar logicalaccuracy gain (%p): +7.07 at L4-L8; negative at L22-L26Mixed
split rest group into L-sinks and ordinary tokens (3-group) (ablation)oracle key gating · CVBench Relationaccuracy gain (%p)No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy