https://smartchunks.com/fluidpd-in-place-elasticity-prefill-decode-llm-serving/Smart Chunksen2026-10-08T09:03:30-04:00FluidPD Lets LLM Servers Reassign GPU Roles Mid-Flight to Dodge Latency Violations https://smartchunks.com/nvidia-nemotron-3-gold-ioi-imo-2026/Smart Chunksen2026-10-07T09:04:44-04:00One Model Family, Two Gold-Level Results: Fine-Tuning Nemotron for IOI and IMO