jackyk02/qwen3.6-35b-a3b-fp8-single-token-embeddings

Qwen3.6-35B-A3B-FP8 single-token embeddings The embedding of every one-token sequence [token_id] under Qwen/Qwen3.6-35B-A3B-FP8 (vLLM pooling runner, convert="embed", last-token pooling, L2-normalized), for all 248,077 token ids of its tokenizer. Why it exists. vLLM's pooling of a one-token prompt returns noise for this model: the last-token embedding of a length-1 prompt has cosine ~0 to the correct vector, alone, batched or mixed with longer prompts, with prefix caching on or… See the full description on the dataset page: https://huggingface.co/datasets/jackyk02/qwen3.6-35b-a3b-fp8-single-to

<iframe src="https://silytics.si/embed/item/81f56edf-e30c-4488-9370-653c55cc3df3" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 23 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.