arXiv:2606. 06906v1 Announce Type: cross Abstract: Long-context question answering (QA) remains challenging for smaller language models even when answer-bearing evidence is already present in the input.
arXiv:2608. 01672v1 Announce Type: cross Abstract: Effective long-context modeling is not merely about retaining more of the past, but about preserving the information that may prove relevant later.
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.
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.
arXiv:2607. 09415v1 Announce Type: cross Abstract: Long-context processing has become increasingly important for large language models (LLMs), but simply extending the context window does not guarantee effective utilization of long inputs.
arXiv:2607. 19358v1 Announce Type: new Abstract: Recent advances in long chain-of-thought reasoning models such as DeepSeek-R1 have led to increasingly longer inference context lengths under the test-time scaling paradigm.
arXiv:2606. 06197v1 Announce Type: cross Abstract: Question answering (QA) systems have achieved notable progress with the advent of large language models (LLMs).
arXiv:2605. 16928v2 Announce Type: replace-cross Abstract: Long-context inference in large language models is bottlenecked by the quadratic cost of full attention.
The paper introduces TreeFuse, a training‑free framework that builds a semantic hierarchy from a full document and uses it to select and organize evidence for a query. By combining structural–semantic scoring with lexical attribution, TreeFuse retains supporting facts and relations while focusing the context on the question. Experiments on four large‑context models show a 1.72‑point improvement in QA‑F1 over full‑context prompting and a 58.55% reduction in answer‑context length.
arXiv:2607. 22690v1 Announce Type: new Abstract: Long-term memory lets LLM agents reuse past interactions, but raw dialogue histories are verbose and information-sparse.
arXiv:2609.13141v1 Announce Type: new Abstract: Post-training attention sparsification reduces the quadratic cumulative attention cost of pretrained Transformers by selecting a small set of context u...
arXiv:2606. 28876v1 Announce Type: cross Abstract: Long-context language models often conflate two different goals: compressing history into an efficient state, and maintaining reliable long-term memory.