A coral-nursery log only makes sense if a bleaching patch in the photo is tied to a phrase in the diver note before any decision. Which fusion type should the team pick?
Select an answer to reveal the explanation.
Short Explanation
When the patch must point at the phrase before anyone decides, you need a mid-network conversation. That is intermediate fusion—not a late handshake or a blind early glue.
Full Explanation
If a coral-nursery log only makes sense when a bleaching patch in the photo is tied to a phrase in the diver note before any decision, the team should pick intermediate fusion so streams interact before the output. Late fusion that only votes after each side decides in isolation never binds patch to phrase mid-network. Early raw concat that skips learned encoders and never lets streams interact mid-network fails the required grounding. Running text alone and ignoring the bleaching patch drops the visual evidence. Exam caveat: required cross-modal grounding before the decision points to intermediate fusion. Operational check: add mid-level interaction (for example cross-attention) between nursery vision and note codes; verify decisions change when the patch-phrase link is ablated. Ablate patch–phrase interaction layers and confirm coral decisions degrade, proving intermediate need. Choose intermediate fusion when bleaching patches must bind to diver phrases before the coral decision. Prove coral intermediate fusion by ablating patch–phrase interaction and showing decision quality drops when streams cannot meet mid-network. Close q123 with a municipal evidence note that records why the correct mechanism fits this civic multimodal scenario, why each of the three distractors fails by a distinct concept, and which operational check was observed in logs or dataset cards before promotion. File that note with the run packet so later maintenance cannot silently revive a rejected framing.