Challenge: Existing methods to reduce sequence length rely on heuristics that break compatibility with hardware-efficient kernels like FlashAttention.
Approach: They propose a method that selectively halts stabilized tokens by monitoring layer-wise update dynamics of the self-attention mechanism.
Outcome: The proposed method can reduce prefill complexity while preserving model accuracy and hardware efficiency.

Similar Papers

DELTA: Dynamic Layer-Aware Token Attention for Efficient Long-Context Reasoning (2026.findings-acl)

Copied to clipboard

Challenge: Large reasoning models generate long chains of intermediate steps, but their inference cost is dominated by decoding, where each new token must attend to the entire growing sequence.
Approach: They propose a training-free sparse attention mechanism that reduces inference cost by evicting entries from the key-value cache.
Outcome: The proposed model matches or surpasses full attention on reasoning benchmarks . it reduces the number of attended tokens by up to 4.25 and delivers 1.54 speedup .
Focus-dLLM: Accelerating Long-Context Diffusion LLM Inference via Confidence-Guided Context Focusing (2026.acl-long)

Copied to clipboard

Challenge: Existing methods for estimating attention importance for tokens are ineffective . dLLMs require bidirectional attention, which limits inference efficiency .
Approach: They propose a training-free attention sparsification framework for efficient long-context inference . they propose 'sink-aware pruning strategy' to accurately estimate and remove redundant computation .
Outcome: The proposed approach offers 29 lossless speedup under 32K context length.
Reducing Token Redundancy in LVLMs: A Systematic Review of Token Pruning Methods (2026.acl-long)

Copied to clipboard

Challenge: Large Vision-Language Models (LVLMs) excel at visual understanding but face severe computational bottlenecks when processing high-resolution images and long videos due to massive visual token counts.
Approach: They propose a taxonomy categorizing methods into vision-side, LLM-side and hybrid paradigms and analyze token selection mechanisms and pruning strategy.
Outcome: The proposed method selectively removes less informative tokens while maintaining performance.
Discovering the Gems in Early Layers: Accelerating Long-Context LLMs with 1000x Input Token Reduction (2026.findings-acl)

Copied to clipboard

Challenge: Large Language Models (LLMs) have demonstrated remarkable capabilities in handling long context inputs, but this comes at the cost of increased computational resources and latency.
Approach: They propose an algorithm that uses early LLM layers as filters to select and compress input tokens, reducing the context length for subsequent processing.
Outcome: The proposed method outperforms existing techniques on the Needle in a Haystack task while demonstrating comparable performance on the LongBench challenge.
POP: Prefill-Only Pruning for Efficient Large Model Inference (2026.findings-acl)

Copied to clipboard

Challenge: Existing structured pruning methods suffer from significant accuracy degradation . Existing pruning methods are expensive and require specialized hardware and kernels to perform .
Approach: They propose a stage-agnostic pruning approach that overlooks asymmetric roles between prefill and decode stages.
Outcome: The proposed pruning approach achieves 1.37 speedup in prefill latency with minimal performance loss.
Stable Language Model Pre-training by Reducing Embedding Variability (2024.emnlp-main)

Copied to clipboard

Challenge: Stable pre-training is essential for achieving better-performing language models, but tracking pre-train stability is impractical due to high computational costs.
Approach: They propose to use Token Embedding Variability as a proxy to estimate pre-training stability.
Outcome: The proposed method improves stability and lowers perplexities even at deeper layer counts.
Vista-LLM: Decoupled Query-Guided Visual Token Pruning for Efficient Long-Video Large Language Models (2026.acl-long)

Copied to clipboard

Challenge: Long-video understanding is bottlenecked by the high cost of processing massive visual tokens.
Approach: They propose a decoupled framework for query-guided visual token pruning . their method reduces visual tokens by 90% and accelerates inference by 98% .
Outcome: The proposed framework reduces visual tokens by 90% and accelerates inference while retaining over 98% of baseline performance on average.
Evolving Sparsity: Leveraging Token Importance Dynamics for Efficient LLM Decoding with Sparse Attention (2026.acl-long)

Copied to clipboard

Challenge: Efficient long-context inference remains a major challenge for large language models (LLMs), as the cost of attention computation during auto-regressive decoding grows linearly with the context length.
Approach: They propose to model token importance as a dynamic process that evolves over decoding steps and propagates through model layers.
Outcome: The proposed method outperforms baseline sparse attention methods and achieves speedups of up to 5.36 for attention latency and 2.33 for end-to-end decoding.
Squeezed Attention: Accelerating Long Context Length LLM Inference (2025.acl-long)

Copied to clipboard

Challenge: Emerging Large Language Models require long input context to perform complex tasks.
Approach: They propose an algorithm to reduce the complexity of attention with respect to the fixed context length.
Outcome: The proposed method reduces the complexity of attention from linear to logarithmic with respect to the fixed context length.
SepSeq: A Training-Free Framework for Long Numerical Sequence Processing in LLMs (2026.findings-acl)

Copied to clipboard

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.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations