Quantization, distillation, pruning and serving work aimed at the same accuracy for less memory, latency and money.
arXiv:2509.04719v3 Announce Type: replace-cross Abstract: The widespread adoption of diffusion models for image generation necessitates efficient parallel inference to manage their substantial comput...
arXiv:2608.14835v2 Announce Type: replace Abstract: Dynamic scene graphs (DSGs) capture spatio-temporal interactions across videos as $\langle$subject, predicate, object$\rangle$ triplets, and underp...
SlideGen is a collaborative vision‑language multi‑agent framework designed to generate scientific presentation slides from research papers. It assigns specialized agents to outline the presentation structure, align figures and tables with key claims, generate speaker notes, and compose editable PPTX slides using a diverse layout library. The system introduces a geometry‑aware density metric to evaluate visual clutter and demonstrates significant improvements in layout balance, content coverage, and text coherence over existing baselines on a 200‑paper benchmark.
arXiv:2608.22188v1 Announce Type: cross Abstract: Split Federated Fine-tuning (SFF) is a promising paradigm for scaling Large Language Models (LLMs) by partitioning model depth between resource-const...
The paper introduces a training‑free structured suffix modeling technique to accelerate Diffusion Language Models (DLMs). It partitions the suffix into local, middle, and tail regions, retaining varying numbers of tokens per region and incorporating previous decoding results into current token representations. Experiments on three DLMs show significant speedups—up to 72.81× in long‑sequence inference—while often improving performance, and the method is compatible with existing acceleration strategies.
arXiv:2608.22898v1 Announce Type: new Abstract: Diffusion language models (DLMs) alleviate the inherent latency bottleneck of autoregressive (AR) large language models (LLMs), but their degraded gene...
arXiv:2608.22344v1 Announce Type: new Abstract: 3D Gaussian Splatting (3DGS) achieves state-of-the-art rendering quality at real-time speeds but suffers from "model bloat" - a large number of redunda...
The paper introduces DRY, a sampling-time logit adjustment that penalizes token generation only when it would extend the current suffix into an exact repetition of an earlier span, thereby preventing verbatim loops in large language model outputs. Experiments across models ranging from 1.5B to 120B parameters and various prompt families show that DRY cuts suffix-extension rates by 47% and improves lexical diversity, while preserving benchmark performance. The method has been adopted by popular open-source LLM inference frameworks, indicating its practical relevance.
arXiv:2602.13106v2 Announce Type: replace-cross Abstract: In recent years, there has been growing interest in understanding neural architectures' ability to learn to execute discrete algorithms, a li...
arXiv:2606.14782v3 Announce Type: replace-cross Abstract: Multimodal Large Language Models (MLLMs) achieve strong vision-language reasoning but incur large KV caches and high decoding latency with lo...
An interactive world model must follow the user's actions, remember the places it has shown, and stream in real time. The tension is structural: control wants a short horizon, memory wants an unbounde...
Vision-Language-Action (VLA) models can turn multimodal context into robot actions, but their action decoders are still trained largely by behavior cloning. This supervises which motor command was dem...
Structured data exists in many forms (tables, knowledge graphs, charts, and time series), and converting it into text may involve different generation tasks. However, most prior work on data-to-text (...
The article discusses a specialized LLM inference runtime designed for real-time applications, such as a 33 ms robot control cycle. Unlike typical runtimes that ignore physical deadlines, this system refuses new requests when the deadline is at risk, evicts key‑value cache entries based on meaning rather than age, and is implemented entirely in hand‑written CUDA without relying on cuBLAS or libtorch.
Learned KV-cache eviction often faces a soft-to-hard mismatch: during training, differentiable gates typically attenuate token contributions, whereas inference saves memory only when KV entries are ph...
We propose HetSkills, a novel framework designed to progressively learn heterogeneous skills within a unified latent space for physics-based character control. The core idea is to treat this latent sp...
Speculative decoding leverages idle CPU resources to accelerate token generation without altering model outputs. In vLLM benchmarks, DFlash achieved a 3.92× increase in autoregressive throughput using Qwen3.5‑9B on an Intel Xeon 6 at a concurrency of 1. The article details the origins of this speedup, discusses acceptance metrics, and outlines factors that influence when speculation is beneficial.
Diffusion Language Models (DLMs) exhibit strong parallel decoding capabilities by denoising multiple tokens in a single generation step. However, this parallelism comes with substantial computational...
Federated Learning (FL) enables distributed clients to collaboratively train models without sharing raw data, making it promising for leveraging massive devices in communication networks. In distillat...
Image matching is a core component of applications such as Simultaneous Localization and Mapping (SLAM), Visual Localization, and Structure from Motion (SfM). However, the local image features central...