Unverified50% confidenceFactExact time
AROMA decomposes VLM control into five independent questions: where to intervene (where), which geometry and gain to use (which geometry and gain), how to reach the target (how), which target to pursue (which target), and when to act (when).
1
Sources
50%
Confidence
Long-term
Relevance
11/13/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedVGR needs to be manually designed or induced from data, and its quality determines the upper limit of the VLM judgment; if a certain aesthetic dimension is missed, the model will not optimize in that direction.64% similarUnverifiedIn scientific chart question answering, VLMs must complete multi-level subtasks such as OCR, legend recognition, axis reading, numerical comparison, and arithmetic reasoning, and a deviation in any step can cause the final answer to mismatch the ground-truth answer.64% similarUnverifiedThe traditional two-stage approach uses a VLM as the planner, outputting textual instructions that are then handed to a video-motion model for execution, which suffers from information loss and the inability to feed back execution errors.63% similarUnverifiedThe effectiveness of the HnS method is highly dependent on the concept retrieval capability of the VLM and the localization accuracy of concept regions.62% similarVerifiedThe core framework of reinforcement learning consists of five elements—agent, environment, state, action, and reward—built on the Markov Decision Process (MDP) and describing decision problems with the five-tuple (S, A, P, R, γ)62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1165379API
curl https://kongchang.com/api/v1/knowledge/claims/1165379MCP
get_claim(id=1165379)