Knowledge catalogue
arXiv:2607.22562v1 Announce Type: new Abstract: Managing long-context dependencies remains a primary bottleneck in LLM agents, as redundant and irrelevant information can degrade multi-step reasoning.
arXiv:2607.21692v1 Announce Type: cross Abstract: Sparse attention reduces the cost of long contexts by allowing each query to read only selected parts of the input. These selectors are often trained
Written by: Shebin Mathew Introduction The 'Golden SAML' technique, first described by CyberArk researchers in 2017, and further detailed by Mandiant researchers in 2021, remains one of the most effec
arXiv:2506.20699v2 Announce Type: replace Abstract: Learning in non-stationary and multi-context environments requires more than ordinary within-task generalization. A system must also discover which
Accessing enterprise data is shifting from static reports to dynamic use by autonomous systems. To keep up, organizations must route fragmented data from SaaS, IoT, and legacy sources into secure, sca
arXiv:2605.13687v1 Announce Type: cross Abstract: We introduce a family of synthetic languages with hierarchical structure -- generated by a broadcast process on trees -- for which the role of context
arXiv:2605.13188v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly deployed in settings where the available context is incomplete or degraded. We argue that an LLM generat
arXiv:2604.20920v1 Announce Type: new Abstract: Scaling large language models to long contexts is challenging due to the quadratic computational cost of full attention. Mitigation approaches include K
Auto-generated index of all companies mentioned across the wiki.
For once I actually agree with the constructive argument here. There are more approaches that allow you to own your own destiny. Memory is just context after all and you can provide that at many place
arXiv:2610.21375v1 Announce Type: cross Abstract: Memory enables large language models (LLMs) to use information beyond their parameters, such as past user conversations, long documents, or an agent's
arXiv:2604.06138v2 Announce Type: replace-cross Abstract: Long-context audio reasoning is underserved in both training data and evaluation. Existing benchmarks target short-context tasks, and the open
arXiv:2610.18014v1 Announce Type: cross Abstract: Non-prefix KV cache reuse is a promising approach for improving inference efficiency in multi-agent systems as agents share the same context blocks. H
Tool-augmented LLM agents discover relevant APIs from large registries before planning execution, yet finite context windows prohibit injecting all tool definitions into a single prompt. The dominant
arXiv:2610.19723v1 Announce Type: new Abstract: Interpreting an ambiguous utterance in real-world interactions often requires context beyond a single utterance. When the utterance refers to objects in
arXiv:2610.15384v1 Announce Type: new Abstract: Compressing the key-value (KV) cache is essential for serving long-context large language models (LLMs) under memory constraints. While current scalar q
arXiv:2610.14583v1 Announce Type: cross Abstract: Autoregressive robot video world models based on discrete visual tokens remain unreliable over long rollout horizons. Retaining the entire generated h
arXiv:2610.14512v1 Announce Type: new Abstract: Objective. To evaluate whether spatial context and temporal modeling improve tissue recognition during optical trocar entry, where multiple abdominal-wa
Hello everyone! My current setup has an RX 7800 XT 16GB GPU + 32GB RAM. I'm currently running Qwen3.8 27B at 4bit with max context 120k at around 40 t/s but the quality isn't that great, and I'd like
arXiv:2610.13735v1 Announce Type: cross Abstract: Efficient attention mechanisms, such as linear and sliding window attention, aim to reduce the computational complexity of Transformers from quadratic
arXiv:2410.11687v4 Announce Type: replace Abstract: In-context learning lets a sequence model adapt to a new task from examples in its input. A prominent line of work shows how self-attention can be c
arXiv:2610.02953v1 Announce Type: new Abstract: Long-context LLM serving is increasingly bottlenecked by KV-cache memory, especially in resource-constrained scenarios. Among existing KV-cache compress
arXiv:2609.38909v1 Announce Type: cross Abstract: Large language models often know the truth and say otherwise: a model that answers correctly when asked neutrally will affirm a user's mistaken belief
arXiv:2609.35750v1 Announce Type: new Abstract: Scaling the horizon of agentic LLMs is bottlenecked by the need to fit ever longer context traces in GPU memory. Context compaction has been the most po
arXiv:2609.34604v1 Announce Type: new Abstract: World models learn environment dynamics from interaction experience. These dynamics depend on the current state and actions, as well as on properties th
arXiv:2609.30884v1 Announce Type: new Abstract: Large language models rely on KV caching to reduce repeated prefill computation in long context and interactive applications. As lightweight adapters ev
arXiv:2609.30789v1 Announce Type: new Abstract: In low-data structure-activity prediction, the choice of molecular representation can matter more than the choice of predictor, and tabular foundation m
arXiv:2609.30245v1 Announce Type: new Abstract: 3D Gaussian Splatting (3DGS) enables high-quality novel-view synthesis but requires substantial storage. Existing compression methods often rely on spat
arXiv:2609.20804v1 Announce Type: new Abstract: Coding harnesses shape how autonomous coding agents translate model capabilities into long-horizon software-engineering performance, yet existing work t
I'm pretty sure it can be done with any model based on qwen4exp, which Qwen's next local models will be based on. You can use a quant that barely fits in VRAM and still run at the model's maximum cont
The full K2 Horizon lineup is out on Artificial Analysis. The AA intelligence vs. parameters plots show that - 0.9B and 375B are bad - 3.7B and 7B are SOTA - 36B A4B is SOTA for hardware with poor mem
metal : single-source fusion table + fusion debug rework (#28164) metal : rework fusion patterns into a single table All fusable op patterns for the Metal backend are now declared once in a fusion tab
arXiv:2609.11744v1 Announce Type: cross Abstract: Prefix caching can reduce the time to first token (TTFT) of long-context LLM requests by reusing previously computed key-value (KV) states, but for sh
arXiv:2609.09768v1 Announce Type: cross Abstract: In Retrieval-Augmented Generation (RAG) systems, a large number of retrieved chunks are concatenated to form the input context so that users can recei
Nice paper to improve inference efficiency. It's been a while we haven't seen good work on efficiency. Here is why it matters: A long-running agent's workspace outgrows its context window long before
arXiv:2609.04915v1 Announce Type: new Abstract: Many long-horizon LLM deployments face tight prompt budgets: latency, cost, and context limits make full-context prompting impractical as interaction le
https://village-sim-one.vercel.app/ - 16GB VRAM RTX 5070 Ti, fully offloaded - Vision on CPU - Windows, not headless - beellama.cpp - latest version with the kvarn performance enhancements making it a
Overview llama.cpp 0.4.0 adds initial Qwen3.8-Flash-Next and Nemotron-3-Puzzle support, on-demand tensor reading, per-slot server context limits, video input options, and a ggml update to 0.23.0 with
arXiv:2609.01925v1 Announce Type: cross Abstract: The attention prefilling phase of long-context LLM inference scales quadratically, making self-attention a severe computational bottleneck. Traditiona
Continual learning isn't some fake term but a real goal and arguably one of the most important open ones in 2026. It roughly means: ‘building systems that keep acquiring useful knowledge or skills aft
arXiv:2608.30076v1 Announce Type: new Abstract: Single-GPU deployment of 70B-parameter language models on an NVIDIA GPU is constrained by device memory, long-context throughput, and engineering integr
arXiv:2608.29477v1 Announce Type: cross Abstract: Document question-answering systems increasingly answer questions over collections of retrieved documents rather than one clean source, so robustness
arXiv:2608.26949v1 Announce Type: new Abstract: Answering questions over real-world documents requires processing long inputs that interleave text with tables. Optical context compression, which repre
arXiv:2608.25907v1 Announce Type: new Abstract: Driver behavior is heterogeneous, context-dependent, and changes over time, and these properties shape the traffic phenomena we observe. Most models, ho
arXiv:2608.25717v1 Announce Type: new Abstract: Retrieval-augmented generation (RAG) is widely assumed to mitigate factual errors in large language models (LLMs), but it remains unclear whether retrie
arXiv:2608.24063v1 Announce Type: cross Abstract: While Vision Large Language Models (VLLMs) have achieved remarkable success in multimodal reasoning, their long-context inference remains prohibitivel
ollama ps settles this in one line, and almost nobody runs it before changing their quant. The PROCESSOR column reports exactly one of three things: 100% GPU - every layer on the GPU 100% CPU - the GP
TL;DR: I'm planning to spend around 100 on cloud GPUs to benchmark Qwen3.8-27B with a focus on questions that actually matter when running it locally: different quant levels/providers, 8-bit vs 16-bit
Thanks to Unsloth Dynamic 3.0 quants coming in slightly leaner and better preserved, I settled on Qwen 3.8 27B (`UD_Q4_K_XL`) at 100K context as my daily driver for Hermes Agent and OpenCode. On an RT
Hi all Looking at a PowerColor R9700 for Qwen3.8-27B, Q4, llama.cpp/Vulkan. AMD's own blog quotes 51.8 tok/s but doesn't say what context length that's at, or whether MTP=2 was holding up. Separately
arXiv:2608.20335v1 Announce Type: new Abstract: We present 4DAnyone, a framework for reconstructing 4D humans from an uncalibrated monocular video by generating reconstruction-grade multiview-consiste
arXiv:2608.19208v1 Announce Type: new Abstract: Multimodal large language models (MLLMs) are frequently exposed to auxiliary textual context, the impact of which on visually grounded tasks remains und
arXiv:2608.18849v1 Announce Type: new Abstract: Tabular foundation models (TFMs) achieve strong performance through in-context learning, but context-dependent inference imposes substantial latency and
arXiv:2608.17164v1 Announce Type: new Abstract: Textual context such as news, reports, and logs can provide valuable signals for time series forecasting, especially when future dynamics are driven by
Every data team knows the moment. Someone opens a table, sees a column called cust_seg_flg, and has to go ask around to find out what it means, whether it's safe to use, and whether anyone has already
arXiv:2602.21219v2 Announce Type: replace-cross Abstract: Large Language Model (LLM) personalization holds great promise for tailoring responses by leveraging personal context and history. However, re
Original Article - https://x.com/p_nawrot/status/2089315591010079034 I've spent the last few years working on efficient attention and KV Cache Compression. I've read many papers, dug deep into referen
arXiv:2608.14191v1 Announce Type: cross Abstract: The key-value (KV) cache stores information from past tokens and is a major memory bottleneck in long-context inference. Existing quantization methods
Qwen3.8-Max is live on Together AI. Together AI is with us as a Day 0 launch partner, and we couldn’t ask for a better name to share Day 0 with. 2.4T parameters, 95B active, 1M context — all together
arXiv:2608.13385v1 Announce Type: new Abstract: Implicit multimodal in-context learning compresses demonstrations into internal interventions, ranging from static task vectors to query-conditioned tra