hardNewOpenAINVIDIATogether AI

gpt-oss adds a learned sink logit to every attention head. How does that change a fused attention kernel and its split-KV merge?

One extra scalar per head looks like a one-line change. The answer finds the two places a fused kernel gets it wrong, and why the error depends on the batch size.

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

One extra scalar per head looks like a one-line change. The answer finds the two places a fused kernel gets it wrong, and why the error depends on the batch size.

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

Next in this trackWhat does DSA's lightning indexer plus top-2,048 selection cost at 128K context, and when does it beat dense attention?Next in this trackSparse MLA decode on a B200 reads FP8 or FP4 KV but multiplies in BF16. Is it bandwidth-bound, and why dequantize?Next in this trackWhich attention kernel will vLLM and SGLang actually run on an H100 and a B200, and what silently changes the answer?

DISCUSSION · 0

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