Commit Graph

  • f8f0712103 docs: record K3 Round 08 release main wuyang 2026-07-30 18:46:12 +08:00
  • f177fa676d research: publish AttnRes forward training study wuyang 2026-07-30 18:41:18 +08:00
  • 7ea91caabb experiment: implement AttnRes forward training runner wuyang 2026-07-30 15:14:15 +08:00
  • d81aacfc68 research: preregister AttnRes forward training study wuyang 2026-07-30 15:03:31 +08:00
  • badf498598 docs: record K3 round seven release wuyang 2026-07-30 14:43:16 +08:00
  • dc8ec30f78 site: explain K3 round seven local path audit wuyang 2026-07-30 14:35:48 +08:00
  • 2dbfd7944a research: audit AttnRes local path matrix wuyang 2026-07-30 14:35:32 +08:00
  • 39a9ad6215 experiment: implement AttnRes local path runner wuyang 2026-07-30 12:53:22 +08:00
  • 6911efc6e7 research: preregister AttnRes local path study wuyang 2026-07-30 12:46:59 +08:00
  • 5e64990a7d docs: record AttnRes spike release wuyang 2026-07-30 12:29:51 +08:00
  • b39378006a site: surface K3 round six on homepage wuyang 2026-07-30 12:26:35 +08:00
  • a1a52d4280 site: explain AttnRes spike path study wuyang 2026-07-30 12:20:47 +08:00
  • 3cbb163715 research: audit AttnRes spike path study wuyang 2026-07-30 12:03:37 +08:00
  • e9ca86c21e research: lock AttnRes spike analyzer wuyang 2026-07-30 11:58:34 +08:00
  • 54c80fb115 research: lock AttnRes spike diagnostic runner wuyang 2026-07-30 10:54:08 +08:00
  • 204f599304 research: preregister AttnRes spike path study wuyang 2026-07-30 10:44:50 +08:00
  • 97ec50a642 docs: record AttnRes gradient release wuyang 2026-07-30 10:30:38 +08:00
  • f7670efcdd feat: add AttnRes gradient scale lab wuyang 2026-07-30 10:24:42 +08:00
  • 26fc824409 research: audit AttnRes gradient scale study wuyang 2026-07-30 09:49:53 +08:00
  • 3f7cc1f544 research: align AttnRes residual precision wuyang 2026-07-30 07:54:38 +08:00
  • 1675ca54f3 research: lock AttnRes gradient runner wuyang 2026-07-30 07:53:12 +08:00
  • 091a05f0a0 research: preregister AttnRes gradient scale study wuyang 2026-07-30 07:46:07 +08:00
  • 40f6a1d631 docs: record reduced AttnRes release wuyang 2026-07-30 07:35:29 +08:00
  • 4ce780dcc9 feat: add reduced AttnRes trace lab wuyang 2026-07-30 07:30:01 +08:00
  • d1d9d22bf3 research: audit reduced AttnRes study wuyang 2026-07-30 07:15:05 +08:00
  • 5f49906be6 fix: flatten noncontiguous AttnRes targets wuyang 2026-07-30 06:39:29 +08:00
  • 1f20f81939 research: freeze reduced AttnRes corpus wuyang 2026-07-30 06:38:58 +08:00
  • f99872039f research: add reduced AttnRes runner wuyang 2026-07-30 06:38:26 +08:00
  • e3617532de research: lock reduced AttnRes model contract wuyang 2026-07-30 06:34:52 +08:00
  • 9039de1b02 research: preregister reduced AttnRes study wuyang 2026-07-30 06:33:38 +08:00
  • 592599fd59 docs: record task bootstrap CRN release wuyang 2026-07-30 06:21:51 +08:00
  • 975ed3dce2 feat: add task bootstrap CRN lab wuyang 2026-07-30 06:15:45 +08:00
  • f041c15e81 research: audit task bootstrap CRN experiment wuyang 2026-07-30 05:59:38 +08:00
  • 65a91ae178 research: add task bootstrap evaluation pipeline wuyang 2026-07-30 04:45:40 +08:00
  • a09e6b14ce research: freeze Chat prompts and CRN runner wuyang 2026-07-30 04:23:14 +08:00
  • 6181af90c8 research: preregister task bootstrap CRN protocol wuyang 2026-07-30 04:14:58 +08:00
  • 756e0e6b18 docs: record cross-source sampling release wuyang 2026-07-30 04:03:12 +08:00
  • 9211333234 feat: audit DeepSeek Chat across sources wuyang 2026-07-30 03:56:55 +08:00
  • 29ceae4e1b docs: record sampling robustness release wuyang 2026-07-30 02:05:55 +08:00
  • 580f69675c feat: map DeepSeek Chat sampling robustness wuyang 2026-07-30 02:01:50 +08:00
  • 18b16e2fdc docs: record completion depth release wuyang 2026-07-30 00:30:37 +08:00
  • 8bb488f275 feat: trace DeepSeek Chat completion depth wuyang 2026-07-30 00:24:27 +08:00
  • fb44d15bb8 docs: record DeepSeek Chat behavior release wuyang 2026-07-29 22:53:36 +08:00
  • 96443d4dfc feat: add DeepSeek Chat behavior evidence wuyang 2026-07-29 22:47:03 +08:00
  • b615224798 docs: record DeepSeek factorial release wuyang 2026-07-29 21:49:40 +08:00
  • 1d0a331e06 feat: factor DeepSeek boundary and role blocks wuyang 2026-07-29 21:42:07 +08:00
  • 8cebb8cb30 docs: record DeepSeek role marker release wuyang 2026-07-29 20:49:32 +08:00
  • ea49ed0511 feat: isolate DeepSeek role marker head token wuyang 2026-07-29 20:44:43 +08:00
  • 8ef2d33437 docs: record DeepSeek boundary control release wuyang 2026-07-29 20:06:22 +08:00
  • f4d4a05dfa feat: isolate DeepSeek history boundary token wuyang 2026-07-29 20:00:37 +08:00
  • f5a69170c7 docs: record DeepSeek equal-length control release wuyang 2026-07-29 19:06:03 +08:00
  • 8130e3745f feat: add DeepSeek equal-length history control wuyang 2026-07-29 19:01:17 +08:00
  • bdd93a5222 docs: record DeepSeek history-factorial release wuyang 2026-07-29 18:11:11 +08:00
  • efe3ffcb4e feat: add DeepSeek message-history factorial probe wuyang 2026-07-29 18:05:59 +08:00
  • 28f8066626 docs: record DeepSeek template-routing release wuyang 2026-07-29 17:28:32 +08:00
  • b0acc8b47b feat: add DeepSeek chat-template routing probe wuyang 2026-07-29 17:23:19 +08:00
  • 405c816d55 docs: record DeepSeek length-control release wuyang 2026-07-29 16:22:07 +08:00
  • 9ca08501a8 feat: add paired DeepSeek routing length control wuyang 2026-07-29 16:17:18 +08:00
  • 059458f8e3 docs: record DeepSeek routing corpus release wuyang 2026-07-29 15:36:28 +08:00
  • 5bcfd58452 feat: add bootstrapped DeepSeek routing corpus wuyang 2026-07-29 15:32:03 +08:00
  • b7fef2350f docs: record absorbed MLA release wuyang 2026-07-29 15:01:40 +08:00
  • 9ba26da545 feat: execute DeepSeek absorbed MLA cache wuyang 2026-07-29 14:57:06 +08:00
  • ab19381046 docs: record DeepSeek execution release wuyang 2026-07-29 14:24:12 +08:00
  • e86420556f feat: trace DeepSeek V2-Lite real routes wuyang 2026-07-29 14:18:45 +08:00
  • a97653af69 docs: record FlashKDA execution release wuyang 2026-07-29 13:42:17 +08:00
  • 2ef846f751 feat: execute FlashKDA on RTX 5090 wuyang 2026-07-29 13:36:49 +08:00
  • 1146208b5e docs: record K3 artifact release wuyang 2026-07-29 12:49:28 +08:00
  • be2b291a04 feat: audit K3 open model artifacts wuyang 2026-07-29 12:44:37 +08:00
  • c3d332680c docs: record K3 second-round release wuyang 2026-07-29 12:08:33 +08:00
  • b669615225 feat: deepen Kimi K3 report guide wuyang 2026-07-29 12:02:04 +08:00
  • 6560e97737 docs: record DeepSeek second-round release wuyang 2026-07-29 11:22:09 +08:00
  • cd96dab22d feat: deepen DeepSeek technical lineage wuyang 2026-07-29 11:17:53 +08:00
  • 5c8a2e44bd docs: record representation release evidence wuyang 2026-07-29 10:37:39 +08:00
  • a2c9298d35 feat: publish representation and depth chapter wuyang 2026-07-29 10:32:33 +08:00
  • a9afa12c7c docs: record evaluation safety production release wuyang 2026-07-29 09:43:44 +08:00
  • cfd2367e88 feat: add evaluation and safety chapter wuyang 2026-07-29 09:33:59 +08:00
  • 1683e19b16 docs: record inference serving production release wuyang 2026-07-29 08:43:26 +08:00
  • db2b7d3570 feat: add inference serving chapter wuyang 2026-07-29 08:39:32 +08:00
  • a7492e1a59 docs: record multimodal production release wuyang 2026-07-29 07:55:32 +08:00
  • 9f5673213d feat: publish native multimodal chapter wuyang 2026-07-29 07:51:42 +08:00
  • 8e5f08ac21 docs: record agent production release wuyang 2026-07-29 06:59:31 +08:00
  • 25982cbbb4 feat: publish long-horizon agent chapter wuyang 2026-07-29 06:54:31 +08:00
  • 6c8c0fb90f docs: record alignment deployment wuyang 2026-07-29 06:11:34 +08:00
  • 8378e2a741 feat: publish alignment deep dive wuyang 2026-07-29 06:07:42 +08:00
  • a030de41fa docs: record transformer deployment wuyang 2026-07-29 05:31:50 +08:00
  • 252e7d6b63 feat: deepen attention and transformer chapter wuyang 2026-07-29 05:26:59 +08:00
  • 266eac8fd6 docs: record language model origins deployment wuyang 2026-07-29 04:52:40 +08:00
  • a54152dc24 feat: publish language model origins chapter wuyang 2026-07-29 04:47:36 +08:00
  • 2997434f7f docs: record numerics deployment wuyang 2026-07-29 04:05:52 +08:00
  • c21af182fb feat: publish numerics and optimization chapter wuyang 2026-07-29 04:02:56 +08:00
  • 2a74cba527 docs: record data engineering deployment wuyang 2026-07-29 03:13:28 +08:00
  • ee0361b3de feat: publish data engineering chapter wuyang 2026-07-29 03:07:40 +08:00
  • 229deecb00 docs: record scaling deployment wuyang 2026-07-29 02:20:48 +08:00
  • eff0e4c942 feat: publish scaling laws chapter wuyang 2026-07-29 02:17:31 +08:00
  • 4c4b5b0c7c research: map llm scaling laws wuyang 2026-07-29 01:41:40 +08:00
  • 3f995101ac docs: record training systems deployment wuyang 2026-07-29 01:32:02 +08:00
  • c6306fe260 feat: publish training systems deep dive wuyang 2026-07-29 01:26:38 +08:00
  • c3e1e85c67 research: map large-scale training systems wuyang 2026-07-29 01:03:15 +08:00
  • 8ae2499951 docs: record reasoning deployment wuyang 2026-07-29 00:43:56 +08:00
  • dce94b283f feat: publish reasoning deep dive wuyang 2026-07-29 00:40:09 +08:00