Kimi researcher says KDA+MLA should use RoPE only if experiments justify it
ChengleiSi · x · 2026-07-29
A Kimi-related researcher says whether KDA+MLA uses RoPE should be decided purely by experimental results, not by organizational context.
- The post explicitly says the choice has nothing to do with whether the author is at Kimi or not
- It also says they continue to use MLA because the experiments support it
- The message is framed as respecting what the data says rather than following assumptions
More from Research
- A former RRE investor breaks down who could win the humanoid robotics race — MarwaEldiwiny · 2026-07-30
- AI Safety Researchers Debate Why Model Goal Guarding Mechanisms Fail — RyanGreenblatt · 2026-07-30
- ITSMBench Released: Frontier Models Struggle with Enterprise Agent Reliability — Shahules786 · 2026-07-30
- Users report Krea-2 LoRA training working in about 2 hours on an RTX 3060 — ganrocks007 · 2026-07-30
- Harvey Open-Sources Legal Agent Benchmark Spanning 1,200 Tasks — baseten · 2026-07-30
- DeepMind’s AI co-scientist solved a 10-year bacterial gene-transfer problem in two days — nathanbenaich · 2026-07-30