arXiv AI September 28, 2026 By Mingqian Yu, Wenpeng Zhang, Shaobo Cui, Peilin Zhao

T-LoopFormer: Token-Level Elastic-Depth Looped Transformers for Latent Reasoning with Dynamic Routing

T-LoopFormer introduces token-level elastic-depth looped transformers that allow each token to decide its own number of loop iterations based on its hidden state, improving token generation accuracy. It also adds a recursion-wise key‑value cache so tokens at different depths only attend to their corresponding cached states, speeding up autoregressive decoding. Experiments demonstrate strong performance on language modeling and zero‑shot reasoning, achieving the lowest decoding latency among comparable models.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.

arXiv AI

Sep 15

T-LoopFormer: Token-Level Elastic-Depth Looped Transformers for Latent Reasoning With Dynamic Routing

arXiv:2609.15160v1 Announce Type: new Abstract: Looped Transformers have recently demonstrated strong performance in both reasoning and language tasks by reusing a shared set of parameters across mul...

By Mingqian Yu, Wenpeng Zhang, Peilin Zhao

llms diffusion efficiency benchmarks

arXiv Machine Learning

Aug 31

Think-at-Hard: Dynamic Looped Transformers for Improved Reasoning

arXiv:2511.08577v4 Announce Type: replace-cross Abstract: Improving the reasoning abilities of Large Language Models (LLMs), especially under parameter constraints, is crucial for real-world applicat...

By Tianyu Fu, Yichen You, Zekai Chen, Guohao Dai, Huazhong Yang, Yu Wang

llms fine-tuning benchmarks

arXiv Computation and Language

Sep 24

Distilling Sequential Computation in Transformer Language Models

arXiv:2609.27233v1 Announce Type: new Abstract: Transformer language models process sequences token by token in an autoregressive manner, making growing contexts increasingly expensive. Yet many adja...

By Zixuan Lan, Jessica Yang, Yanhong Li, Karen Livescu, Jiawei Zhou

llms rag nlp efficiency

arXiv AI

Jun 16

Think-at-Hard: Selective Latent Iterations to Improve Reasoning Language Models

arXiv:2511. 08577v3 Announce Type: replace-cross Abstract: Improving the reasoning abilities of Large Language Models (LLMs), especially under parameter constraints, is crucial for real-world applications.

By Tianyu Fu, Yichen You, Zekai Chen, Guohao Dai, Huazhong Yang, Yu Wang

llms fine-tuning benchmarks

arXiv AI

Jul 17

T^2MLR: Transformer with Temporal Middle-Layer Recurrence

arXiv:2607. 15178v1 Announce Type: cross Abstract: Transformer reasoning is limited by autoregressive decoding, which repeat edly compresses rich hidden computation through token space and makes it difficult for intermediate reasoning states to persist across time.

By Ziyang Cai, Xingyu Zhu, Yihe Dong, Yinghui He, Sanjeev Arora

llms fine-tuning

arXiv AI

Aug 11

Full-bandwidth transformer

arXiv:2608. 08888v1 Announce Type: new Abstract: Autoregressive transformers compute along two axes: horizontally across generated tokens, and vertically through model depth.

By Xi Wang, Ziyang Cai, Zheng Zhan, Harry Dong, Ying Fan, Gustavo de Rosa, Tim Pearce, John Langford

llms rag efficiency

One email a morning, machine-written

Related stories

T-LoopFormer: Token-Level Elastic-Depth Looped Transformers for Latent Reasoning With Dynamic Routing

Think-at-Hard: Dynamic Looped Transformers for Improved Reasoning

Distilling Sequential Computation in Transformer Language Models

Think-at-Hard: Selective Latent Iterations to Improve Reasoning Language Models

T^2MLR: Transformer with Temporal Middle-Layer Recurrence

Full-bandwidth transformer