Disentangle to Decay: Linear Attention with Trainable Decay Factor (2025.coling-main)
Copied to clipboard
| Challenge: | Existing linear attention models use a decay factor based positional encoding (PE), but the decay factor is manually designed and non-trainable, limiting further optimization. |
| Approach: | They propose a PE-based positional encoding that disentangles decay factor into two parts to achieve further optimization and stable training. |
| Outcome: | The proposed model achieves stable training of decay factor and improves inference efficiency in normal context and extrapolation scenarios. |
Similar Papers
Fine-Tuning Pre-trained Transformers into Decaying Fast Weights (2022.emnlp-main)
Copied to clipboard
| Challenge: | Autoregressive Transformers incur O(T) complexity during per-token generation due to the self-attention mechanism. |
| Approach: | They propose a kernel-based method to approximate causal self-attention by replacing it with recurrent formulations with various update rules and feature maps to achieve O(1) time and memory complexity. |
| Outcome: | The proposed method outperforms prior methods and retains 99% of attention’s performance on WikiText-103 against more complex attention substitutes. |
Modeling Context With Linear Attention for Scalable Document-Level Translation (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Document-level machine translation models lack quadratic complexity in the sequence length due to their attention layers. |
| Approach: | They evaluate a recent linear attention model with a sentential gate to promote a recency inductive bias and compare it to open-source document translation. |
| Outcome: | The proposed model significantly improves translation quality on IWSLT 2015 and OpenSubtitles 2018 with similar or better BLEU scores. |
Sparse Attention with Linear Units (2021.emnlp-main)
Copied to clipboard
| Challenge: | Recent studies have suggested that sparse attention mechanisms can be made more interpretable by replacing the softmax activation with its sparser variants. |
| Approach: | They propose a method to replace softmax activation with a ReLU to achieve sparsity in attention by layer normalization with either a specialized initialization or an additional gating function. |
| Outcome: | The proposed model is easy to implement and more efficient than previously proposed sparse attention mechanisms. |
Position-Aware Depth Decay Decoding (D3): Boosting Large Language Model Inference Efficiency (2025.findings-acl)
Copied to clipboard
| Challenge: | Recent dynamic computation methods show that not all components are required for inference, enabling a training-free pipeline. |
| Approach: | They propose a token-position aware layer skipping framework to save 1.5x times operations efficiently while maintaining performance. |
| Outcome: | The proposed algorithm achieves 1.5x speedup on large language models with no retraining and with comparable performance on the GSM8K and BBH benchmarks. |
Attention Entropy is a Key Factor: An Analysis of Parallel Context Encoding with Full-attention-based Pre-trained Language Models (2025.acl-long)
Copied to clipboard
Zhisong Zhang, Yan Wang, Xinting Huang, Tianqing Fang, Hongming Zhang, Chenlong Deng, Shuaiyi Li, Dong Yu
| Challenge: | Large language models have demonstrated remarkable performance across a wide range of language tasks due to their remarkable ability in context modeling. |
| Approach: | They propose to use parallel context encoding to reduce attention entropy by incorporating attention sinks and selective mechanisms to reduce irregular attention . they also propose to incorporate attention sink mechanisms into the parallel encoded context to reduce the irregular attention. |
| Outcome: | The proposed methods lower irregular attention entropy and narrow performance gaps. |
Dynamic Attention-Guided Context Decoding for Mitigating Context Faithfulness Hallucinations in Large Language Models (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing methods, such as a n-terminal coding, do not provide accurate data for large language models. |
| Approach: | They propose a lightweight framework that leverages attention distributions and uncertainty signals in a single-pass decoding. |
| Outcome: | Experiments on open-book QA datasets show that DAGCD improves faithfulness and robustness while preserving computational efficiency. |
Conv-Basis: A New Paradigm for Efficient Attention Inference and Gradient Computation in Transformers (2025.findings-emnlp)
Copied to clipboard
| Challenge: | a large computational cost for attention computation in large language models is a major obstacle . |
| Approach: | They propose a convolution-like structure for attention computation using convolution matrices . they then propose an efficient approximation method to approximate the attention matrix . |
| Outcome: | The proposed method achieves nearly linear time complexity in n1+o(1) time. |
PDR: A Plug-and-Play Positional Decay Framework for LLM Pre-training Data Detection (2026.acl-long)
Copied to clipboard
| Challenge: | Existing likelihood-based methods for detecting pretraining data are limited in black-box, zero-shot settings. |
| Approach: | They propose a training-free and plug-and-play framework that reweights token-level scores to amplify distinct signals from early positions while suppressing noise from later ones. |
| Outcome: | The proposed framework amplifys signals from early positions while suppressing noise from later positions. |
Increasing Learning Efficiency of Self-Attention Networks through Direct Position Interactions, Learnable Temperature, and Convoluted Attention (2020.coling-main)
Copied to clipboard
| Challenge: | SANs are an integral part of successful neural networks such as Transformer . training SAN on a task or pretraining them on language modeling requires large amounts of data and compute resources. |
| Approach: | They propose to modify SANs to enable faster learning, i.e., higher accuracies after fewer update steps. |
| Outcome: | The proposed modifications enable faster learning, i.e., higher accuracies after fewer update steps. |
ReGLA: Refining Gated Linear Attention (2025.naacl-long)
Copied to clipboard
| Challenge: | Recent advances in Large Language Models (LLMs) are known for their computational and storage requirements due to the quadratic computation complexity of softmax attention. |
| Approach: | They propose to reduce the quadratic computation complexity of softmax attention by using feature maps, normalization and the gating mechanism to improve performance. |
| Outcome: | The proposed model outperforms existing gated linear attention models in extensive tasks including training from scratch and post-linearization with continual pre-training. |