Unverified50% confidenceFactExact time
The LRM (Large Reconstruction Model) architecture was proposed by Hong Kong University and Adobe Research in 2023, reframing 3D reconstruction as a sequence-to-sequence prediction task.
1
Sources
50%
Confidence
Long-term
Relevance
8/16/2026
First Seen
Sources
Modly：本地GPU运行的开源图像转3D建模工具
Related Claims
Unverified现代LLM-based Agent流行架构由Yao等人在2023年的ReAct论文中奠基，将推理与行动交织在同一生成过程中62% similarUnverifiedLLM作为调度中枢的设计模式被称为ReAct（Reasoning + Acting）框架61% similarUnverified模型复现是根据论文描述重新实现并验证机器学习模型的过程，是ML工程中最考验综合能力的任务之一60% similarUnverifiedDPM-Solver由Cheng Lu等人于2022年提出，将扩散模型反向ODE重写为半线性结构，采样步数压缩至约10步60% similarUnverifiedLLMs are generative AI models based on the Transformer architecture and pretrained on massive text data, with representative products including OpenAI's GPT, Anthropic's Claude, and Meta's LLaMA.59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/757008API
curl https://kongchang.com/api/v1/knowledge/claims/757008MCP
get_claim(id=757008)