tracked author

Xin Cheng

First author and arXiv submitter of Engram / Conditional Memory via Scalable Lookup, and equal-contribution contact author of DSpark, with Peking University and DeepSeek-AI affiliation on both papers. Public profile evidence beyond paper/project pages is sparse, so no personal X account is recorded.

2 archived notes X: not-found

Related Notes

按论文归档时间排序,展示该作者在本站已经出现的材料。

Jun 28, 2026

DSpark: Confidence Scheduled Speculative Decoding with Semi Autoregressive Generation

DSpark 是 DeepSeek 把并行 drafter 推向生产 serving 的一套完整方案:用 DFlash 式 parallel backbone 先一次生成长候选块,再用低秩 Markov head 注入块内局部自回归依赖,随后用 calibrated confidence head 和硬件感知 prefix scheduler 按请求与负载动态裁剪 target verification 长度。离线 Qwen3 / G...

2026-06-27-dspark-confidence-scheduled-speculative-decoding Speculative DecodingMulti-Token PredictionServing Runtime
Jun 24, 2026

Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models

Engram 把大模型稀疏性从 MoE (Mixture of Experts,混合专家,用条件计算扩大 total 参数并控制 active compute) 扩展到 conditional memory:用 hashed N gram lookup 存静态局部模式,用 context aware gate 决定是否把查到的记忆注入 hidden state;在 iso parameter / iso FLOPs 的 27B MoE...

2601.07372-conditional-memory-engram-scalable-lookup Memory ArchitectureMoE Architecture