MoNe is a lightweight modular neural memory that can be attached to any frozen pretrained Transformer to enable efficient long‑context inference without retraining. It processes context in fixed‑size segments using test‑time learning of fast‑weight neural memory networks with layer‑localized gradient updates, and during inference the memory generates keys and values from query tokens alone, avoiding re‑reading context tokens. This two‑phase design decouples inference cost from context length, achieving linear preprocessing and constant query cost while keeping peak GPU memory independent of context size; at 128K tokens it cuts compute and memory usage by about 80% with only a 6.4% parameter overhead, and it performs well on long‑context benchmarks where in‑context learning degrades.
arXiv:2609.25537v1 Announce Type: new Abstract: Large language model (LLM) inference is constrained by the quadratic scaling of self-attention and the linear scaling of the KV cache, increasing laten...
arXiv:2603. 29002v3 Announce Type: replace-cross Abstract: Modern large language models (LLMs) increasingly depends on efficient long-context processing and generation mechanisms, including sparse attention, retrieval-augmented generation (RAG), and compressed contextual memory, to support complex reasoning.
Elastic Threshold Attention (ETA) is a trainable attention mechanism that dynamically predicts contextual thresholds from query representations, enabling selective pruning of KV cache tokens during long‑context decoding. By multiplicatively suppressing sub‑threshold logits during training, ETA avoids representation collapse and eliminates localized attention sinks, allowing a 1.45B model to match dense attention performance at roughly 85% training sparsity and 38% active decode density. At inference, a custom Triton kernel achieves up to 2.5× faster decoding on sequences up to 512K tokens, and an offline calibration step can further reduce compute by 27% by freezing per‑head thresholds.
The paper introduces Ladder Side Tuning (LST), a parameter‑efficient fine‑tuning method that adds a lightweight side network to large language models. LST matches QLoRA’s compute scaling while halving peak memory usage, enabling 7B‑parameter models to be fine‑tuned on a single 12 GB GPU with 2k‑token contexts without gradient checkpointing. The authors also present xLadder, a depth‑extended variant that increases effective depth through cross‑connections, allowing deeper reasoning without extra memory overhead.
arXiv:2606. 09079v1 Announce Type: cross Abstract: Conventional LLMs keep the full KV cache loaded during decoding, causing a severe GPU memory bottleneck for ultra-long context serving.
arXiv:2602. 21196v2 Announce Type: replace Abstract: Efficiently processing long sequences with Transformer models usually requires splitting the computations across accelerators via context parallelism.
arXiv:2606. 09659v1 Announce Type: cross Abstract: Long-context language model inference is bottlenecked by memory, as the KV cache grows with context length.
Long-context language model inference is bottlenecked by memory, as the KV cache grows with context length. Recent techniques to compress the KV cache fall short: they either degrade model quality substantially or require considerable time and compute to compress a single long prompt.
The paper introduces the Tri‑Metric Router, a deterministic, training‑free policy that chooses among Raw, Neural, and Lexical pipelines for retrieval‑augmented generation on commodity GPUs. It uses three CPU‑side signals—spatial complexity, syntactic density, and type‑token ratio—to balance VRAM headroom and latency, calibrated on LongBench qasper. The method eliminates out‑of‑memory failures and improves alignment and F1 scores compared to always‑on lexical compression without extra VRAM or training costs.
arXiv:2608. 19920v1 Announce Type: new Abstract: A lot of prior work addressed key-value (KV) cache selection and compression by sparse attention to enable long-context inference for transformer language models without excessive hardware budgets.
arXiv:2606. 16093v1 Announce Type: cross Abstract: Modeling long-range dependencies remains a central challenge in natural language processing.
A lot of prior work addressed key-value (KV) cache selection and compression by sparse attention to enable long-context inference for transformer language models without excessive hardware budgets. We provide a new method for fine-tuning models with sparse attention.