arXiv:2607. 01241v1 Announce Type: cross Abstract: Existing prompt compression methods treat text as flat token sequences, failing to capture the distributed nature of important information, which is often spread across multiple locations and connected through both local syntactic dependencies and global semantic relations.
arXiv:2505. 23277v3 Announce Type: replace-cross Abstract: Retrieval-augmented generation (RAG) often suffers from long and noisy retrieved contexts.
CommunityKV is a new framework that treats sparse attention as a community detection problem, building a token graph from $QK^T$ scores and partitioning it into semantically coherent communities. It updates token communities in constant time during streaming decoding, avoiding costly global re‑partitioning. Experiments on Qwen3 and Llama‑3.1 show that CommunityKV can increase end‑to‑end generation throughput by up to 1.25×, and with query‑group graph aggregation up to 1.71×, while maintaining comparable accuracy.
arXiv:2603.09222v2 Announce Type: replace Abstract: Efficient context compression is critical for retrieval-augmented question answering in resource-constrained settings, where long retrieved context...
arXiv:2607. 17486v1 Announce Type: cross Abstract: As large language models (LLMs) process increasingly longer prompts, computation and KV-cache memory costs have emerged as major bottlenecks in inference systems.
arXiv:2607. 02980v1 Announce Type: cross Abstract: Scaling modern large language models (LLMs) to long contexts is limited by the quadratic computation cost, and poor length extrapolation of dense attention.
The paper introduces SCSP, a training‑free framework that improves long‑context embeddings by selectively pooling informative tokens. SCSP partitions documents into sentence‑aware chunks, adds a semantic compression prompt to each chunk, and uses prompt‑isolated attention masks to estimate token importance. The selected tokens’ intermediate‑layer representations are aggregated to form the final embedding, yielding consistent performance gains across zero‑shot and fine‑tuned models on long‑context benchmarks.
arXiv:2608. 08569v1 Announce Type: new Abstract: Recent advancements in Speech Large Language Models have demonstrated remarkable capabilities in understanding complex audio tasks.
arXiv:2609.37052v1 Announce Type: new Abstract: Omnimodal large language models (Omni-LLMs) encode audio and visual streams into temporally interleaved token sequences for multimodal reasoning. Howev...
arXiv:2606. 09659v1 Announce Type: cross Abstract: Long-context language model inference is bottlenecked by memory, as the KV cache grows with context length.
CAGE: Coherence-Aware Graph Encoding for Retrieval-Augmented Generation introduces a reranking framework that evaluates and enhances the coherence of retrieved passages across four dimensions—Intra-Domain Relevance, Noise Resistance, Informational Bonding, and Factual Consistency. The method transforms passages into directed heterogeneous entity graphs, reweights factual anchors, encodes structural patterns with a Relational Graph Convolutional Network, and fuses inter-chunk coherence with query relevance to produce a final ranking. Evaluations on four multi‑hop benchmarks show that CAGE matches or surpasses strong baselines, improving Recall@5 on bridge‑dominated datasets and consistently boosting downstream Exact Match scores, indicating that structurally coherent context leads to more precise answers even when retrieval recall is similar or lower.
Large language models (LLMs) have demonstrated rapidly improving long-context capabilities, prompting a wave of benchmarks designed to evaluate them. However, existing long-context evaluations - from Needle-in-a-Haystack (NIAH) tests to more recent multi-hop reasoning and summarization tasks - predominantly measure average-case performance, and many are either saturated or lack robustness.