Commit Graph

44 Commits

Author SHA1 Message Date
wuyang f041c15e81 research: audit task bootstrap CRN experiment 2026-07-30 05:59:38 +08:00
wuyang a09e6b14ce research: freeze Chat prompts and CRN runner 2026-07-30 04:23:14 +08:00
wuyang 6181af90c8 research: preregister task bootstrap CRN protocol 2026-07-30 04:14:58 +08:00
wuyang 9211333234 feat: audit DeepSeek Chat across sources 2026-07-30 03:57:17 +08:00
wuyang 580f69675c feat: map DeepSeek Chat sampling robustness 2026-07-30 02:01:50 +08:00
wuyang 8bb488f275 feat: trace DeepSeek Chat completion depth 2026-07-30 00:24:27 +08:00
wuyang 96443d4dfc feat: add DeepSeek Chat behavior evidence 2026-07-29 22:47:49 +08:00
wuyang 1d0a331e06 feat: factor DeepSeek boundary and role blocks 2026-07-29 21:42:39 +08:00
wuyang ea49ed0511 feat: isolate DeepSeek role marker head token 2026-07-29 20:44:43 +08:00
wuyang f4d4a05dfa feat: isolate DeepSeek history boundary token 2026-07-29 20:00:37 +08:00
wuyang 8130e3745f feat: add DeepSeek equal-length history control 2026-07-29 19:01:17 +08:00
wuyang efe3ffcb4e feat: add DeepSeek message-history factorial probe 2026-07-29 18:05:59 +08:00
wuyang b0acc8b47b feat: add DeepSeek chat-template routing probe 2026-07-29 17:23:19 +08:00
wuyang 9ca08501a8 feat: add paired DeepSeek routing length control 2026-07-29 16:17:18 +08:00
wuyang 5bcfd58452 feat: add bootstrapped DeepSeek routing corpus 2026-07-29 15:32:03 +08:00
wuyang 9ba26da545 feat: execute DeepSeek absorbed MLA cache 2026-07-29 14:57:06 +08:00
wuyang e86420556f feat: trace DeepSeek V2-Lite real routes 2026-07-29 14:18:45 +08:00
wuyang a97653af69 docs: record FlashKDA execution release 2026-07-29 13:42:17 +08:00
wuyang 2ef846f751 feat: execute FlashKDA on RTX 5090 2026-07-29 13:36:49 +08:00
wuyang 1146208b5e docs: record K3 artifact release 2026-07-29 12:49:28 +08:00
wuyang be2b291a04 feat: audit K3 open model artifacts 2026-07-29 12:44:37 +08:00
wuyang c3d332680c docs: record K3 second-round release 2026-07-29 12:08:33 +08:00
wuyang b669615225 feat: deepen Kimi K3 report guide 2026-07-29 12:02:04 +08:00
wuyang cd96dab22d feat: deepen DeepSeek technical lineage 2026-07-29 11:17:53 +08:00
wuyang a2c9298d35 feat: publish representation and depth chapter 2026-07-29 10:32:33 +08:00
wuyang a9afa12c7c docs: record evaluation safety production release 2026-07-29 09:43:44 +08:00
wuyang cfd2367e88 feat: add evaluation and safety chapter 2026-07-29 09:33:59 +08:00
wuyang db2b7d3570 feat: add inference serving chapter 2026-07-29 08:39:32 +08:00
wuyang 9f5673213d feat: publish native multimodal chapter 2026-07-29 07:52:27 +08:00
wuyang 25982cbbb4 feat: publish long-horizon agent chapter 2026-07-29 06:54:31 +08:00
wuyang 8378e2a741 feat: publish alignment deep dive 2026-07-29 06:07:42 +08:00
wuyang 252e7d6b63 feat: deepen attention and transformer chapter 2026-07-29 05:27:38 +08:00
wuyang a54152dc24 feat: publish language model origins chapter 2026-07-29 04:47:36 +08:00
wuyang c21af182fb feat: publish numerics and optimization chapter 2026-07-29 04:02:56 +08:00
wuyang ee0361b3de feat: publish data engineering chapter 2026-07-29 03:07:40 +08:00
wuyang 4c4b5b0c7c research: map llm scaling laws 2026-07-29 01:41:40 +08:00
wuyang 3f995101ac docs: record training systems deployment 2026-07-29 01:32:02 +08:00
wuyang c6306fe260 feat: publish training systems deep dive 2026-07-29 01:26:38 +08:00
wuyang c3e1e85c67 research: map large-scale training systems 2026-07-29 01:03:15 +08:00
wuyang 8ae2499951 docs: record reasoning deployment 2026-07-29 00:43:56 +08:00
wuyang dce94b283f feat: publish reasoning deep dive 2026-07-29 00:40:09 +08:00
wuyang 3db826dd4a feat: publish MoE deep dive 2026-07-28 23:32:14 +08:00
wuyang 761a75d5af feat: publish long-context deep dive 2026-07-28 22:49:04 +08:00
wuyang 7d5c5e4c54 feat: launch LLM Atlas research course 2026-07-28 21:55:32 +08:00