Expired50% confidenceEventExact time
A Reddit user shared an experiment using Qwen 3.6 VLM to find Wally in a dense crowd scene from Where's Wally, with results showing the model struggled with high-density target search tasks.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
8/28/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 9/11/2026(expired)
Sources
Related Entities
Related Claims
UnverifiedVLM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64% similarVerifiedTraditional RLHF methods have a fundamental limitation of 'reward hacking,' where models may learn to perform well in evaluation scenarios but fail in out-of-distribution scenarios60% similarUnverifiedQDAM 3.7 Plus's weaknesses include a tendency toward excessive thematic elevation, thin period-specific details, and insufficient web novel feel.60% similarUnverifiedImageNet竞赛曾暴露研究者通过反复调优超参数刷榜、导致模型基准表现与实际部署效果出现显著差距的问题59% similarUnverifiedQwen 3.8 Max may not perform stably enough during long-running tasks, with occasional issues such as failing to create files when it should during longer sessions.58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/810988API
curl https://kongchang.com/api/v1/knowledge/claims/810988MCP
get_claim(id=810988)