inference efficiency

From Long to Lean: Performance-aware and Adaptive Chain-of-Thought Compression via Multi-round Refinement EMNLP 2025 From PEFT to DEFT: Parameter Efficient Finetuning for Reducing Activation Density in Transformers AAAI 2025 Generative Prompt Internalization NAACL 2025 Graph Sparsification via Mixture of Graphs ICLR 2025 HiRED: Attention-Guided Token Dropping for Efficient Inference of High-Resolution Vision-Language Models AAAI 2025 HShare: Fast LLM Decoding by Hierarchical Key-Value Sharing ICLR 2025 IG-Pruning: Input-Guided Block Pruning for Large Language Models EMNLP 2025 IIET: Efficient Numerical Transformer via Implicit Iterative Euler Method EMNLP 2025 Improving Prediction Certainty Estimation for Reliable Early Exiting via Null Space Projection IJCAI 2025 KV-Latent: Dimensional-level KV Cache Reduction with Frequency-aware Rotary Positional Embedding ACL 2025 Large Language Models Are Overparameterized Text Encoders NAACL 2025 LightThinker: Thinking Step-by-Step Compression EMNLP 2025 Long Context Compression with Activation Beacon ICLR 2025 Lookahead Q-Cache: Achieving More Consistent KV Cache Eviction via Pseudo Query EMNLP 2025 LVPruning: An Effective yet Simple Language-Guided Vision Token Pruning Approach for Multi-modal Large Language Models NAACL 2025 Mamba-Shedder: Post-Transformer Compression for Efficient Selective Structured State Space Models NAACL 2025 MaTVLM: Hybrid Mamba-Transformer for Efficient Vision-Language Modeling ICCV 2025 Mitigating Geometric Degradation in Fast DownSampling via FastAdapter for Point Cloud Segmentation ICCV 2025 MLWQ: Efficient Small Language Model Deployment via Multi-Level Weight Quantization EMNLP 2025 MoDification: Mixture of Depths Made Easy NAACL 2025 More Tokens, Lower Precision: Towards the Optimal Token-Precision Trade-off in KV Cache Compression EMNLP 2025 MrT5: Dynamic Token Merging for Efficient Byte-level Language Models ICLR 2025 Offloaded Reasoning: Efficient Inference for Large Language Models via Modular Reasoning and Refinement EMNLP 2025 OmniCache: A Trajectory-Oriented Global Perspective on Training-Free Cache Reuse for Diffusion Transformer Models ICCV 2025 Palu: KV-Cache Compression with Low-Rank Projection ICLR 2025