The Recurrent Transformer: Greater Effective Depth and Efficient Decoding

• URL: http://arxiv.org/abs/2604.21215v1
• Date: Thu, 23 Apr 2026 02:12:58 GMT
• Title: The Recurrent Transformer: Greater Effective Depth and Efficient Decoding
• Abstract summary: We introduce the Recurrent Transformer, a simple architectural change where each layer attends to key-value pairs computed off its own activations.<n>We show that the architecture can emulate both (i) a conventional Transformer and (ii) token-to-gressive recurrent updates under mild assumptions.
• Score: 48.9323408950142
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Transformers process tokens in parallel but are temporally shallow: at position $t$, each layer attends to key-value pairs computed based on the previous layer, yielding a depth capped by the number of layers. Recurrent models offer unbounded temporal depth but suffer from optimization instability and historically underutilize modern accelerators. We introduce the Recurrent Transformer, a simple architectural change where each layer attends to key-value pairs computed off its own activations, yielding layerwise recurrent memory while preserving standard autoregressive decoding cost. We show that the architecture can emulate both (i) a conventional Transformer and (ii) token-to-token recurrent updates under mild assumptions, while avoiding optimization instability. Naively, prefill/training appears bandwidth-bound with effective arithmetic intensity near $1$ because keys and values are revealed sequentially; we give an exact tiling-based algorithm that preserves the mathematical computation while reducing HBM traffic from $Θ(N^2)$ to $Θ(N\log N)$, increasing effective arithmetic intensity to $Θ(N/\log N)$ for sequence length $N$. On 150M and 300M parameter C4 pretraining, Recurrent Transformers improve cross-entropy over a parameter-matched Transformer baseline and achieve the improvement with fewer layers (fixed parameters), suggesting that recurrence can trade depth for width, thus reducing KV cache memory footprint and inference latency.

Related papers

• T^2MLR: Transformer with Temporal Middle-Layer Recurrence [28.818755729513793]
  We develop a transformers-based latent reasoning architecture that fuses a cached middle layer representation from the previous token directly into an earlier layer of the current token position.<n>Applying recurrence to only a localized middle-layer block (as little as 20% of the network) often outperforms full-layer recurrence.<n>These results suggest that effective latent reasoning in Transformers does not require looping over all layers as in previous works, but can instead emerge more strongly from targeted middle-layer recurrence.
  arXiv  Detail & Related papers  (2026-07-16T16:33:59Z)
• Hidden Decoding at Scale: Latent Computation Scaling for Large Language Models [75.80275843320511]
  We study whether an existing backbone can keep improving by allocating more computation to each token while leaving the Transformer backbone fixed.<n>We propose Hidden Decoding, a sequence-length scaling method applied during continued pretraining.
  arXiv  Detail & Related papers  (2026-07-09T07:37:59Z)
• Latent Recurrent Transformer: Architecture Exploration, Training Strategies, and Scaling Behavior [107.2098567818173]
  Latent Recurrent Transformer (LRT) is a lightweight augmentation of autoregressive transformers.<n>LRT reuses a high-level source-layer hidden state from the previous token as recurrent memory for the next token.
  arXiv  Detail & Related papers  (2026-05-26T10:10:26Z)
• N-vium: Mixture-of-Exits Transformer for Accelerated Exact Generation [68.47358899451255]
  N-vium is a mixture-of-exits transformer that partially parallelizes computation across depth on standard hardware.<n>N-vium attaches prediction heads at multiple depths and defines the next-token distribution as a learned mixture over these exits.
  arXiv  Detail & Related papers  (2026-05-13T08:46:17Z)
• SCORE: Replacing Layer Stacking with Contractive Recurrent Depth [0.0]
  Residual connections are central to modern deep neural networks.<n>We propose SCORE (Skip-Connection ODE Recurrent Embedding), a discrete recurrent alternative to classical layer stacking.
  arXiv  Detail & Related papers  (2026-03-11T08:49:03Z)
• Hyperparameter Transfer Laws for Non-Recurrent Multi-Path Neural Networks [10.793433622094959]
  We introduce a graph-based notion of effective depth for multi-path neural networks.<n>We show that the optimal learning rate decays with effective depth following a universal -3/2 power law.<n>Experiments confirm the predicted slope and enable reliable zero-shot transfer of learning rates across depths and widths.
  arXiv  Detail & Related papers  (2026-02-07T11:10:54Z)
• DeepCoT: Deep Continual Transformers for Real-Time Inference on Data Streams [63.27233749591346]
  Transformer-based models have dramatically increased their size and parameter count to tackle increasingly complex tasks.<n>Stream data inference is typically performed over a sliding temporal window, leading to highly redundant computations.<n>We propose the Deep Continual Transformer (DeepCoT), a redundancy-free encoder-only model that can be applied over existing deep encoder architectures with minimal changes.
  arXiv  Detail & Related papers  (2025-11-21T16:15:43Z)
• Tensor Train Multiplication [0.0]
  The computational complexity and memory requirements of the TTM algorithm scale as $chi3$ and $chi2$, respectively. This represents a significant improvement compared with the conventional approach. The TTM algorithm paves the way towards GPU accelerated tensor network simulations of computational fluid dynamics problems with large bond dimensions.
  arXiv  Detail & Related papers  (2024-10-10T12:36:49Z)
• Chain of Thought Empowers Transformers to Solve Inherently Serial Problems [57.58801785642868]
  Chain of thought (CoT) is a highly effective method to improve the accuracy of large language models (LLMs) on arithmetics and symbolic reasoning tasks. This work provides a theoretical understanding of the power of CoT for decoder-only transformers through the lens of expressiveness.
  arXiv  Detail & Related papers  (2024-02-20T10:11:03Z)
• TCNCA: Temporal Convolution Network with Chunked Attention for Scalable Sequence Processing [52.64837396100988]
  MEGA is a recent transformer-based architecture, which utilizes a linear recurrent operator whose parallel computation, based on the FFT, scales as $O(LlogL)$, with $L$ being the sequence length. We build upon their approach by replacing the linear recurrence with a special temporal convolutional network which permits larger receptive field size with shallower networks, and reduces the computational complexity to $O(L)$. We evaluate TCNCA on EnWik8 language modeling, long-range-arena (LRA) sequence classification, as well as a synthetic reasoning benchmark associative recall.
  arXiv  Detail & Related papers  (2023-12-09T16:12:25Z)
• DeepPCR: Parallelizing Sequential Operations in Neural Networks [4.241834259165193]
  We introduce DeepPCR, a novel algorithm which parallelizes typically sequential operations in order to speed up inference and training of neural networks. DeepPCR is based on interpreting a sequence of $L$ steps as the solution of a specific system of equations, which we recover using the Parallel Cyclic Reduction algorithm. To verify the theoretical lower complexity of the algorithm, and to identify regimes for speedup, we test the effectiveness of DeepPCR in parallelizing the forward and backward pass in multi-layer perceptrons.
  arXiv  Detail & Related papers  (2023-09-28T10:15:30Z)
• Stable, Fast and Accurate: Kernelized Attention with Relative Positional Encoding [63.539333383965726]
  We propose a novel way to accelerate attention calculation for Transformers with relative positional encoding (RPE) Based upon the observation that relative positional encoding forms a Toeplitz matrix, we mathematically show that kernelized attention with RPE can be calculated efficiently using Fast Fourier Transform (FFT)
  arXiv  Detail & Related papers  (2021-06-23T17:51:26Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.