Inner Thinking Transformer: Leveraging Dynamic Depth Scaling to Foster Adaptive Internal Thinking

• URL: http://arxiv.org/abs/2502.13842v1
• Date: Wed, 19 Feb 2025 16:02:23 GMT
• Title: Inner Thinking Transformer: Leveraging Dynamic Depth Scaling to Foster Adaptive Internal Thinking
• Authors: Yilong Chen, Junyuan Shang, Zhenyu Zhang, Yanxi Xie, Jiawei Sheng, Tingwen Liu, Shuohuan Wang, Yu Sun, Hua Wu, Haifeng Wang,
• Abstract summary: We propose Inner Thinking Transformer, which reimagines layer computations as implicit thinking steps. ITT achieves 96.5% performance of a 466M Transformer using only 162M parameters, reduces training data by 43.2%, and outperforms Transformer/Loop variants in 11 benchmarks.
• Score: 51.154226183713405
• License:
• Abstract: Large language models (LLMs) face inherent performance bottlenecks under parameter constraints, particularly in processing critical tokens that demand complex reasoning. Empirical analysis reveals challenging tokens induce abrupt gradient spikes across layers, exposing architectural stress points in standard Transformers. Building on this insight, we propose Inner Thinking Transformer (ITT), which reimagines layer computations as implicit thinking steps. ITT dynamically allocates computation through Adaptive Token Routing, iteratively refines representations via Residual Thinking Connections, and distinguishes reasoning phases using Thinking Step Encoding. ITT enables deeper processing of critical tokens without parameter expansion. Evaluations across 162M-466M parameter models show ITT achieves 96.5\% performance of a 466M Transformer using only 162M parameters, reduces training data by 43.2\%, and outperforms Transformer/Loop variants in 11 benchmarks. By enabling elastic computation allocation during inference, ITT balances performance and efficiency through architecture-aware optimization of implicit thinking pathways.

Related papers

• Transforming Indoor Localization: Advanced Transformer Architecture for NLOS Dominated Wireless Environments with Distributed Sensors [7.630782404476683]
  We introduce a novel tokenization approach, referred to as Sensor Snapshot Tokenization (SST), which preserves variable-specific representations of power delay profile ( PDP) We also propose a lightweight Swish-Gated Linear Unit-based Transformer (L-SwiGLU Transformer) model, designed to reduce computational complexity without compromising localization accuracy.
  arXiv  Detail & Related papers  (2025-01-14T01:16:30Z)
• PredFormer: Transformers Are Effective Spatial-Temporal Predictive Learners [65.93130697098658]
  This paper proposes PredFormer, a pure transformer-based framework for predictive learning. With its recurrent-free, transformer-based design, PredFormer is both simple and efficient. experiments on synthetic and real-world datasets demonstrate that PredFormer achieves state-the-art performance.
  arXiv  Detail & Related papers  (2024-10-07T03:52:06Z)
• UIO-LLMs: Unbiased Incremental Optimization for Long-Context LLMs [111.12010207132204]
  UIO-LLMs is an incremental optimization approach for memory-enhanced transformers under long-context settings. We refine the training process using the Truncated Backpropagation Through Time (TBPTT) algorithm. UIO-LLMs successfully handle long context, such as extending the context window of Llama2-7b-chat from 4K to 100K tokens with minimal 2% additional parameters.
  arXiv  Detail & Related papers  (2024-06-26T08:44:36Z)
• Sparse Universal Transformer [64.78045820484299]
  The Universal Transformer (UT) is a variant of the Transformer that shares parameters across its layers. This paper proposes the Sparse Universal Transformer (SUT), which leverages Sparse Mixture of Experts (SMoE) and a new stick-breaking-based dynamic halting mechanism.
  arXiv  Detail & Related papers  (2023-10-11T00:38:57Z)
• Optimizing Non-Autoregressive Transformers with Contrastive Learning [74.46714706658517]
  Non-autoregressive Transformers (NATs) reduce the inference latency of Autoregressive Transformers (ATs) by predicting words all at once rather than in sequential order. In this paper, we propose to ease the difficulty of modality learning via sampling from the model distribution instead of the data distribution.
  arXiv  Detail & Related papers  (2023-05-23T04:20:13Z)
• Memristive Stochastic Computing for Deep Learning Parameter Optimization [1.6344851071810071]
  Computing (SC) is a computing paradigm that allows for the low-cost and low-power of various arithmetic operations using bit streams and digital logic. We demonstrate that in using a 40-nm Complementary Metal Oxide Semiconductor (CMOS) process our scalable architecture occupies 1.55mm$2$ and consumes approximately 167$mu$W when optimizing parameters of a Convolutional Neural Network (CNN) while it is being trained for a character recognition task, observing no notable reduction in accuracy post-training.
  arXiv  Detail & Related papers  (2021-03-11T07:10:32Z)
• Optimizing Inference Performance of Transformers on CPUs [0.0]
  Transformers-based models (e.g., BERT) power many important Web services, such as search, translation, question-answering, etc. This paper presents an empirical analysis of scalability and performance of inferencing a Transformer-based model on CPUs.
  arXiv  Detail & Related papers  (2021-02-12T17:01:35Z)
• Bayesian Transformer Language Models for Speech Recognition [59.235405107295655]
  State-of-the-art neural language models (LMs) represented by Transformers are highly complex. This paper proposes a full Bayesian learning framework for Transformer LM estimation.
  arXiv  Detail & Related papers  (2021-02-09T10:55:27Z)

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.