{"ID":2825568,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2512.21323","arxiv_id":"2512.21323","title":"Parallel Token Prediction for Language Models","abstract":"Autoregressive decoding in language models is inherently slow, generating only one token per forward pass. We propose Parallel Token Prediction (PTP), a general-purpose framework for predicting multiple tokens in a single model call. PTP moves the source of randomness from post-hoc sampling to random input variables, making future tokens deterministic functions of those inputs and thus jointly predictable in a single forward pass. We prove that a single PTP call can represent arbitrary dependencies between tokens. PTP is trained by distilling an existing model or through inverse autoregressive training without a teacher. Experimentally, PTP achieves a 2.4x speedup on a diverse-task speculative decoding benchmark. We provide code and checkpoints at https://github.com/mandt-lab/ptp.","short_abstract":"Autoregressive decoding in language models is inherently slow, generating only one token per forward pass. We propose Parallel Token Prediction (PTP), a general-purpose framework for predicting multiple tokens in a single model call. PTP moves the source of randomness from post-hoc sampling to random input variables, m...","url_abs":"https://arxiv.org/abs/2512.21323","url_pdf":"https://arxiv.org/pdf/2512.21323v2","authors":"[\"Felix Draxler\",\"Justus Will\",\"Farrin Marouf Sofian\",\"Theofanis Karaletsos\",\"Sameer Singh\",\"Stephan Mandt\"]","published":"2025-12-24T18:46:55Z","proceeding":"cs.CL","tasks":"[\"cs.CL\",\"cs.LG\"]","methods":"[\"Language Model\"]","has_code":false,"code_links":[{"ID":605676,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_id":2825568,"paper_url":"https://arxiv.org/abs/2512.21323","paper_title":"Parallel Token Prediction for Language Models","repo_url":"https://github.com/mandt-lab/ptp","is_official":false,"mentioned_in_paper":false,"mentioned_in_github":true,"github_stars":0}]}