Expired90% confidenceFactTime unknown
In the offline RL phase, Cursor used the GRPO (Group Relative Policy Optimization) algorithm, an RL algorithm variant proposed by DeepSeek.
1
Sources
90%
Confidence
Medium-term (~90 days)
Relevance
5/31/2026
First Seen
Valid until: 8/29/2026(expired)
Sources
Related Entities
Related Claims
UnverifiedCursor injects content into the model context through a retrieval-augmented generation (RAG) pipeline that combines approximate nearest-neighbor retrieval using code semantic vector embeddings with reference tracking via LSP symbol dependency graphs, followed by reranking.64% similarUnverifiedTensorRT能针对CUDA和GPU优化张量运算，进一步提升性能62% similarUnverifiedThe "fast" optimization label implies that the model may have adopted inference acceleration techniques such as FlashAttention, PagedAttention, or dedicated CUDA kernels61% similarUnverifiedTensorRT深度绑定CUDA生态，提供最激进的自动图优化和kernel自动调优，是车载GPU平台的首选61% similarUnverifiedcuDNN is a GPU-accelerated library optimized specifically for deep neural networks60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/6587API
curl https://kongchang.com/api/v1/knowledge/claims/6587MCP
get_claim(id=6587)