Record · corpus.blog/posts/01a08c5a-7869-73a8-bf7f-499c2c87fc4d

Speculative decoding: when LLMs predict their own predictions

undef.dev · published 26 June 2026

A record is what was published and who pointed at it. The text of the post is not held here: read it at the source, or at its Wayback capture.

Read the post at undef.dev Wayback capture

0blogs citing

First cited: —
Most recent: —
Rank this month: not ranked
Links from this post: 20
Words captured: 3,357

Cited by

No blog we hold has cited this post yet.

Links from this post

Link

Host

https://arxiv.org/abs/2302.01318

arxiv.org

https://arxiv.org/abs/2211.17192

arxiv.org

https://colcarroll.github.io/couplings/static/maximal_couplings.html

colcarroll.github.io

https://arxiv.org/abs/2404.19737

arxiv.org

https://arxiv.org/abs/2412.19437

arxiv.org

https://arxiv.org/abs/2401.15077

arxiv.org

https://arxiv.org/abs/2503.01840

arxiv.org

https://arxiv.org/abs/1703.06114

arxiv.org

https://huggingface.co/Qwen/Qwen3.6-27B

huggingface.co

https://z.ai/blog/glm-5.2

z.ai

https://blog.google/innovation-and-ai/technology/developers-tools/multi-token-prediction-gemma-4/

blog.google

https://arxiv.org/abs/2305.09781

arxiv.org

https://arxiv.org/abs/2406.16858

arxiv.org

https://arxiv.org/abs/2401.10774

arxiv.org

GLM-5.2: Built for Long-Horizon Tasks Read at the source ↗

huggingface.co

https://huggingface.co/google/gemma-4-31B-it-assistant/blob/main/config.json

huggingface.co

https://github.com/huggingface/transformers/blob/main/src/transformers/models/gemma4/configuration_gemma4.py

github.com

https://huggingface.co/Qwen

huggingface.co

https://zolotukhin.ai/blog/2026-05-08-why-mtp-heads-are-the-speculative-decode-draft-qwen3-a3b-deserves/

zolotukhin.ai

https://unsloth.ai/docs/models/qwen3.6

unsloth.ai