LAWCAT: Efficient Distillation from Quadratic to Linear Attention with Convolution across Tokens for Long Context Modeling (2025.findings-emnlp)
Copied to clipboard
Zeyu Liu, Souvik Kundu, Lianghao Jiang, Anni Li, Srikanth Ronanki, Sravan Babu Bodapati, Gourav Datta, Peter Anthony Beerel
| Challenge: | a novel linearization framework is proposed to reduce the cost of training transformers from scratch. |
| Approach: | They propose a linear attention framework that integrates pre-trained transformers into a performant linear attention architecture. |
| Outcome: | The proposed framework improves performance on mistral-7B with 1K-length sequences and BABILong benchmarks. |
Similar Papers
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. |
Latent-Condensed Transformer for Efficient Long Context Modeling (2026.acl-long)
Copied to clipboard
| Challenge: | Existing approaches address these bottlenecks separately: Multi-head Latent Attention (MLA) reduces the KV cache by projecting tokens into a low-dimensional latent space, while sparse attention reduces computation. |
| Approach: | They propose a Latent-Condensed Attention mechanism that performs structured context condensation directly within MLA's latent space. |
| Outcome: | The proposed approach reduces KV cache size and attention cost without adding parameters. |
Context Compression for Auto-regressive Transformers with Sentinel Tokens (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing Transformer-based LLMs have limited performance due to complexity of attention module . key-value cache is the major memory footprint and inference latency problem . |
| Approach: | They propose a plug-and-play approach that incrementally compresses token activation into compact ones . they also profile the benefit of context compression on improving the system throughout . |
| Outcome: | The proposed approach reduces memory footprint and inference latency by compressing tokens into compact ones. |
Efficient Transformer Knowledge Distillation: A Performance Review (2023.emnlp-industry)
Copied to clipboard
| Challenge: | Pretrained transformer language models have been gaining popularity in the field of natural language processing . however, there is no study into the intersection of these two fields . |
| Approach: | They propose a method to extract knowledge from transformers to produce high-performing efficient attention models with low costs. |
| Outcome: | The proposed model compression method preserves up to 98.6% of original model performance across short-context tasks and up to 95.8% on long-concept Named Entity Recognition tasks while decreasing inference times by up to 57%. |
Lizard: An Efficient Linearization Framework for Large Language Models (2026.acl-long)
Copied to clipboard
Chien Van Nguyen, Huy Huu Nguyen, Ruiyi Zhang, Hanieh Deilamsalehy, Puneet Mathur, Viet Dac Lai, Haoliang Wang, Jayakumar Subramanian, Ryan A. Rossi, Trung Bui, Nikos Vlassis, Franck Dernoncourt, Thien Huu Nguyen
| Challenge: | Existing linearization frameworks that rely on softmax attention with quadratic time and memory complexity pose significant computational and memory bottlenecks for long-context applications. |
| Approach: | They propose a linearization framework that transforms pretrained Transformer-based Large Language Models into subquadratic architectures that closely approximate softmax attention while preserving model quality. |
| Outcome: | Experiments show that the proposed framework outperforms existing methods by 9.4 - 24.5 points on the 5-shot MMLU benchmark and demonstrating superior associative recall. |
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. |
SepSeq: A Training-Free Framework for Long Numerical Sequence Processing in LLMs (2026.findings-acl)
Copied to clipboard
Jie Sun, Yu Liu, Lu Han, Qiwen Deng, Xiang Shu, Yang Xiao, Lintao Ma, Xingyu Lu, Jun Zhou, Pengfei Liu, Jiancan Wu, Xiang Wang
| Challenge: | Existing large-scale large-context models suffer from performance degradation when processing long numerical sequences. |
| Approach: | They propose a framework to mitigate attention dispersion by strategically inserting separator tokens into the model to recalibrat attention to local segments while preserving global context. |
| Outcome: | The proposed framework improves accuracy and reduces inference token consumption by 16.4% on 9 widely-adopted LLMs. |
CoCA: Fusing Position Embedding with Collinear Constrained Attention in Transformers for Long Context Window Extending (2024.acl-long)
Copied to clipboard
| Challenge: | Existing models that use self-attention and position embedding have anomalous behavior that hinder long context window extrapolation. |
| Approach: | They propose a collinear constraint between Q and K to integrate RoPE and self-attention. |
| Outcome: | The proposed model integrates self-attention and position embedding into LLMs without fine-tuning. |
Octopus: Gated Selective Attention for Memory-Bounded Long-Context Inference in Large Language Models (2026.acl-long)
Copied to clipboard
| Challenge: | Subquadratic architectures rely on aggressive state compression that degrades performance on complex reasoning tasks. |
| Approach: | They propose a framework that confers fixed-memory inference onto pretrained Transformers . they use a learnable module that enforces an adaptive sparsity policy over the context history . |
| Outcome: | The proposed framework outperforms state-of-the-art linearized baselines on the GSM8K benchmark by over 36 points under identical memory constraints. |
The Devil in Linear Transformer (2022.emnlp-main)
Copied to clipboard
| Challenge: | Existing linear transformers suffer from performance degradations on various tasks and corpus. |
| Approach: | They propose a new linear attention that replaces scaling with a normalization to stabilize gradients and confine attention to neighbouring tokens in early layers. |
| Outcome: | The proposed model outperforms vanilla transformers on the long-range arena benchmark while being significantly more space-time efficient. |