Skip to content
← Speculative & Constrained Decoding / 22
hardNewOpenAIAnthropicGoogle DeepMind

An RL rollout engine speculates with an EAGLE head trained on the step-0 policy, and rollouts are no faster now. Why?

A drafter trained on the starting policy goes stale as training moves the target, and the outputs stay correct while the speedup disappears. The answer derives when speculation stops paying at rollout batch sizes, finds the vLLM paths that refresh the target and leave the drafter behind, and picks a fix.

Updated Nov 2026 · Worked answers on inference mechanisms, with calculations, implementation details, and follow-up questions.

A drafter trained on the starting policy goes stale as training moves the target, and the outputs stay correct while the speedup disappears. The answer derives when speculation stops paying at rollout batch sizes, finds the vLLM paths that refresh the target and leave the drafter behind, and picks a fix.

10 answers per topic instead of 5, plus saved progress and bookmarks · costs nothingor unlock all 213 remaining answers · $25

Every worked answer and derivation in all 12 topics, every concept page, the full practice-test bank, and every code drill’s edge-case tests and reference solution.

6 months · One payment · No auto-renewal

UP NEXT ON YOUR JOURNEY
Popular right nowWhy is batch-1 decode on a 70B model bandwidth-bound, and what is its latency floor on two H100s?Popular right nowDecode attention intensity is fixed at the query-to-KV head ratio. Derive that, then tell me what MLA does to it.Popular right nowAccount for every gigabyte on a GPU running vLLM: weights, KV pool, activations, CUDA graphs and NCCL buffers.
DISCUSSION · 0

Nobody has commented on this answer. Add a correction or another way to work it.