Unverified50% confidenceFactExact time
Many VLMs struggle with tasks such as OCR, chart understanding, and precise counting.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/12/2026
First Seen
Valid until: 12/11/2026
Sources
Related Entities
Related Claims
UnverifiedIn hard sample detection, VLMs have stronger semantic capabilities but at a high computational cost, while lightweight detection models are fast but struggle with hard samples, forming a trade-off between accuracy and efficiency.70% similarUnverifiedVLMs have limitations when processing high-precision engineering drawings: insufficient accuracy in recognizing small text, a tendency to hallucinate and fabricate dimensions, and limited understanding of GD&T professional symbols.69% similarUnverifiedVLM在描述整体场景时表现出色，却在精确定位微小目标时频频出错，这是因为当画面元素密集时单个patch可能包含多个人物导致细节信息在下采样中被稀释或丢失66% similarUnverifiedWhen endowing existing VLMs with generative capabilities, the understanding abilities accumulated during model pretraining are often impaired, resulting in catastrophic forgetting.64% similarUnverifiedThe article's authors conclude that the CLIP encoder is one of the bottlenecks in modern VLMs, but not the only factor, and this issue remains unsettled.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/914596API
curl https://kongchang.com/api/v1/knowledge/claims/914596MCP
get_claim(id=914596)