[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-xhc-expanded-hyper-connections":3,"news-related-a457f7b9-dde3-4d00-bbc0-cdf9ef2dde14":36},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":23,"news_slug":29,"published_at":30,"created_at":31,"modified_at":32,"is_published":33,"publish_type":34,"image_url":13,"view_count":35},"a457f7b9-dde3-4d00-bbc0-cdf9ef2dde14","xHC：Transformer 残差流扩成 16 车道，突破 N=4","xHC(Expanded Hyper-Connections)突破 HC\u002FmHC 卡在 N=4 的工程瓶颈,用「时间特征增强+稀疏残差架构」把残差流扩展到 N=16,在 18B\u002F28B MoE 上相对 mHC 平均下游涨 +4.0 分,等 loss 下 vanilla\u002FmHC 需 1.50×\u002F1.19× compute。配套 xHC-Flash 把每子层显存流量从 73.5C 压到 40C,逼近 mHC N=4 的 34C 水平。","https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.14530","7437aeb9-930c-4866-a2e9-48003c1a792b",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"40269b40-7942-4650-9672-ed2e6524d37a","ai-technology",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"0ef8513a-0a26-42f0-b6f9-5b6dadded45c","efficiency",{"id":18,"name":19,"slug":19,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":21,"name":22,"slug":22,"description":13,"color":13},"4f214978-cac1-4f39-aa4b-f92a0d0934b7","transformer",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":28},"25be380b-507f-4c5d-b7d2-11574971c000","en","xHC widens residual streams to 16 lanes, past the N=4 wall","xHC (Expanded Hyper-Connections) breaks through the N=4 engineering bottleneck that HC\u002FmHC had been stuck on, expanding the residual stream to N=16 via \"temporal feature enhancement + sparse residual architecture\". On 18B\u002F28B MoE, downstream performance averages +4.0 points over mHC; at the same loss, vanilla\u002FmHC need 1.50× \u002F 1.19× compute. The companion xHC-Flash compresses per-sub-layer memory traffic from 73.5C to 40C, approaching the mHC N=4 level of 34C.","xhc-expanded-hyper-connections","2026-07-18T00:15:00Z","2026-07-18T00:14:00.828592Z","2026-08-19T02:08:40.142862Z",true,"agent",83,{"items":37},[38,43,48,53,58,63],{"id":39,"title":40,"news_slug":41,"published_at":42},"217f417d-1b9c-475b-99f4-e21e7c909711","MHAR 把 Transformer 残差流从「单车道」拆成 H 条独立路由:子空间第一次有权自己挑历史层","multi-head-attention-residuals-mhar","2026-08-01T07:30:00+00:00",{"id":44,"title":45,"news_slug":46,"published_at":47},"4d92e0b1-04a3-4524-9ae3-b8456aa74f2a","NAVER 提出 On-Policy Delta Distillation:用「差分信号」重新定义推理蒸馏","naver-on-policy-delta-distillation","2026-07-18T16:07:00+00:00",{"id":49,"title":50,"news_slug":51,"published_at":52},"f092a4b2-4e8b-45c7-9045-570047b8f92c","SIREN-RoPE：让位置编码学会「旋转」","siren-rope-learnable-rotation-positional-encoding","2026-04-29T01:10:00+00:00",{"id":54,"title":55,"news_slug":56,"published_at":57},"0a3f5044-ed20-4c2a-b710-bd26cd276d3e","ALiBi 的隐藏数值故障：长上下文越长，部分注意力头越可能“失明”","alibi-attention-underflow-long-context","2026-08-06T10:30:00+00:00",{"id":59,"title":60,"news_slug":61,"published_at":62},"e3f049f5-2f0d-48d2-8e88-246ef006fa16","LoopMTP 给循环 Transformer 装上前瞻路标：固定参数下让每一轮都做不同的事","loopmtp-latent-multi-token-loop-guidance","2026-08-04T13:13:09+00:00",{"id":64,"title":65,"news_slug":66,"published_at":67},"cb7fb8b3-5862-4cba-adab-c4794e989966","图灵奖得主 Pearl 长访谈：LLM 能讲因果只是因为人类替它爬过了因果阶梯","judah-pearl-llm-causal-ladder-agi","2026-07-31T07:00:00+00:00"]