Parallel Decoder Transformer: Planner-Seeded Latent Coordination for Synchronized Parallel Decoding

TL;DR

The paper introduces the Parallel Decoder Transformer (PDT), a novel architecture enabling synchronized parallel decoding in language models through internal coordination mechanisms, improving upon external prompt-based methods.

Contribution

It presents a frozen-trunk transformer architecture with a planner-seeded latent workspace and synchronized multi-stream output protocol for better parallel decoding coordination.

Findings

Enables synchronized parallel decoding within a frozen language model.

Uses a planner and latent workspace for internal coordination.

Improves parallel task decomposition over external prompting.

Abstract

Autoregressive language models can often identify parallel subproblems, but standard decoding exposes only a single left-to-right output interface. External orchestration methods can launch multiple prompts concurrently, yet they provide no model-internal state through which those generations can synchronize, resolve ownership, or wait for missing information. We present the Parallel Decoder Transformer (PDT), a frozen-trunk architecture that augments a decoder with a planner-seeded latent workspace and a synchronized multi-stream output protocol. Before any stream emits tokens, a mandatory prompt-time planner predicts fixed latent plan slots and projects them as snapshot 0 on an embeddings-only Dynamic Notes Bus. During decoding, each stream reads the visible notes window through Speculative Note Conditioning (SNC), emits provisional token blocks and latent summaries, and advances only when agreement logic determines that the current shared state is sufficient for continued parallel generation. Coverage heads track plan-item ownership, while rollback handles incoherent or premature commits. PDT therefore shifts parallel task decomposition from an external prompting strategy to a model-internal coordination mechanism over the output interface of a frozen language model.