The Context-Ready Transformer

Researchers introduced the context-ready transformer, a recurrent neural network architecture that pre-contextualizes tokens before they enter a transformer block, achieving faster inference and competitive performance. A 5-layer model outperformed a 12-layer standard transformer while generating 1.7x faster on an A100, and a single-layer model with K=10 surpassed a 6-layer transformer with 2.6x speedup. The architecture excels with wide representations and long contexts, solving complex pointer-chasing tasks that standard transformers struggle with.

arXiv:2606.27538v1 Announce Type: new Abstract: We introduce the context-ready transformer, a new recurrent neural network architecture built from a D-layer transformer block that pre-contextualizes each token before it enters the block. During left-to-right generation, a correction network combines the previous position's block output -- a cached summary of past context -- with the current token embedding, so the tokenenters the block already contextualized rather than as a raw embedding. At sequential inference, the correction chain makes the architecture a recurrent neural network. For training, we unroll the correction process K times over the full sequence, processing all positions in parallel at each step. A pretrained transformer can also be converted to a context-ready model by adding a zero-initialized correction FFN and fine-tuning. We evaluate across widths, depths, block sizes, and two datasets, with all comparisons against standard transformers, variants, and ablations. A D=5 model beats a 12-layer transformer while generating 1.7x faster on an A100. With K=10, a single-layermodel D=1 beats a 6-layer transformer with a 2.6x inference speedup, and sequential inference matches parallel K=10 to within 0.01 PPL. The architecture benefits most from wide representations and long contexts. On a pointer-chasing task, D=1 trained with BPTT solves all 10 composition levels, while standard transformers exhibit staircase-like depth dependence.