All entries

RouteLMT: Learned Sample Routing for Hybrid LLM Translation Deployment

arXiv:2604.22520v1 Announce Type: new Abstract: Large Language Models (LLMs) have achieved remarkable performance in Machine Translation (MT), but deploying them at scale remains prohibitively expensi

researcharxiv-cs-cl

RTX 5090 users: TensorRT-LLM vs llama.cpp (GGUF) for Coding Agents (Cline/RooCode) – Is the speed worth the VRAM limit?

DGX agent

This post compares TensorRT-LLM and llama.cpp (GGUF) as inference frameworks for running coding agents like Cline and RooCode on RTX 5090 GPUs, examining the tradeoff between inference speed and VRAM

model-releasesr-ollama

27 Apr 2026

Running Qwen3.5-397B-A17B (4bit quants, 177 GB) on two DGX Sparks using llama.cpp with RPC and RDMA:

model-releasesgeorgi-gerganov--x

This post documents a technical demonstration of running the large Qwen3.5-397B-A17B model across distributed hardware using llama.cpp with advanced networking protocols. The approach leverages 4-bit

Safety and innovation are not mutually exclusive: for many companies, especially in high-trust industries, AI’s risks are also a hindrance t…

Safety and innovation are not mutually exclusive: for many companies, especially in high-trust industries, AI’s risks are also a hindrance to adoption. In this op-ed in the @FT, I underscore that Euro

safetyyoshua-bengio--x

Sam Altman cannot be trusted. • The OpenAI board fired him because he was not always honest with them. They said he should not control power…

Sam Altman cannot be trusted. • The OpenAI board fired him because he was not always honest with them. They said he should not control powerful AI. • A major report talked to over 100 people and saw s

safetyelon-musk--x

SAMIDARE: Advanced Tracking-by-Segmentation for Dense Scenarios

arXiv:2604.22162v1 Announce Type: new Abstract: Automated sports analysis demands robust multi-object tracking (MOT), yet segmentation-based methods often struggle with mask errors and ID switches in

researcharxiv-cs-cv

San Francisco public records: in 2024, Salesforce and Airbnb were among the top tech spenders in the city's 10B program, which lets them hire police officers (Paresh Dave/Wired)

Paresh Dave / Wired: San Francisco public records: in 2024, Salesforce and Airbnb were among the top tech spenders in the city's 10B program, which lets them hire police officers — A recent attack on

Scam Altman

model-releaseselon-musk--x

Scam Altman Interesting how it works Elon puts up his own money, rounds up the absolute best AI talent on the planet, leverages every connection he has to secure serious resources, and launches OpenAI

Scam Altman and Greg Stockman stole a charity. Full stop. Greg got tens of billions of stock for himself and Scam got dozens of OpenAI side …

model-releaseselon-musk--x

Scam Altman and Greg Stockman stole a charity. Full stop. Greg got tens of billions of stock for himself and Scam got dozens of OpenAI side deals with a piece of the action for himself, Y Combinator s

Scam Altman didn’t tell the OpenAI board that he OWNED the OpenAI Startup Fund. Altman lied in congressional testimony that he didn’t have f…

Scam Altman didn’t tell the OpenAI board that he OWNED the OpenAI Startup Fund. Altman lied in congressional testimony that he didn’t have financial gain from OpenAI. Ex-board member of OpenAI calls S

safetyelon-musk--x

Scam Altman owned the OpenAI Startup fund while simultaneously lying to the world that he didn’t financially benefit from OpenAI

Scam Altman owned the OpenAI Startup fund while simultaneously lying to the world that he didn’t financially benefit from OpenAI The main difference between Elon Musk and Scam Altman: Elon Musk actual

industryelon-musk--x

Score-based Membership Inference on Diffusion Models

arXiv:2509.25003v2 Announce Type: replace-cross Abstract: Membership inference attacks (MIAs) against Diffusion Models (DMs) raise pressing privacy concerns by revealing whether a sample was part of t

Applications

Segment Any-Quality Images with Generative Latent Space Enhancement

arXiv:2503.12507v3 Announce Type: replace Abstract: Despite their success, Segment Anything Models (SAMs) experience significant performance drops on severely degraded, low-quality images, limiting th

applicationsarxiv-cs-cv

Segmentation of Gray Matters and White Matters from Brain MRI data

arXiv:2603.29171v3 Announce Type: replace Abstract: Accurate segmentation of brain tissues such as gray matter and white matter from magnetic resonance imaging is essential for studying brain anatomy,

researcharxiv-cs-cv

Selective Contrastive Learning For Gloss Free Sign Language Translation

arXiv:2604.22374v1 Announce Type: new Abstract: Sign language translation (SLT) converts continuous sign videos into spoken-language text, yet it remains challenging due to the intrinsic modality mism

safetyarxiv-cs-cl

Selective Depthwise Separable Convolution for Lightweight Joint Source-Channel Coding in Wireless Image Transmission

arXiv:2604.22338v1 Announce Type: cross Abstract: Depthwise separable convolutional (DSConv) layers have been successfully applied to deep learning (DL)-based joint source-channel coding (JSCC) scheme

Selective Rotary Position Embedding

arXiv:2511.17388v2 Announce Type: replace Abstract: Position information is essential for language modeling. In softmax transformers, Rotary Position Embeddings (extit{RoPE}) encode positions through

researcharxiv-cs-cl

Self-Supervised Multisensory Pretraining for Contact-Rich Robot Reinforcement Learning

arXiv:2511.14427v3 Announce Type: replace-cross Abstract: Effective contact-rich manipulation requires robots to synergistically leverage vision, force, and proprioception. However, Reinforcement Lear

safetyarxiv-cs-lg

Semantic Error Correction and Decoding for Short Block Channel Codes

arXiv:2604.22269v1 Announce Type: cross Abstract: This paper presents a semantic-enhanced receiver framework for transmitting natural language sentences over noisy wireless channels using multiple sho

Sensory-Aware Sequential Recommendation via Review-Distilled Representations

arXiv:2603.02709v2 Announce Type: replace-cross Abstract: We propose a novel framework for sensory-aware sequential recommendation that enriches item representations with linguistically extracted sens

Sereact raises $110M to scale AI ‘robotic brain’ and expand into the US

German robotics startup Sereact GmbH announced today that it has raised 110 million in new funding to scale up its artificial intelligence “robotic brain” platform and expand into the U.S. Founded in

industrysiliconangle

Shaken or Stirred? An Analysis of MetaFormer's Token Mixing for Medical Imaging

arXiv:2510.05971v3 Announce Type: replace Abstract: The generalization of the Transformer architecture via MetaFormer has reshaped our understanding of its success in computer vision. By replacing sel

SHAPE: Unifying Safety, Helpfulness and Pedagogy for Educational LLMs

model-releasesarxiv-cs-cl

arXiv:2604.22134v1 Announce Type: new Abstract: Large Language Models (LLMs) have been widely explored in educational scenarios. We identify a critical vulnerability in current educational LLMs, pedag

Shard the Gradient, Scale the Model: Serverless Federated Aggregation via Gradient Partitioning

arXiv:2604.22072v1 Announce Type: cross Abstract: Federated learning (FL) aggregation on serverless platforms faces a hard scalability ceiling: existing architectures (lambda-FL, LIFL) partition clien

Applications

Shared Lexical Task Representations Explain Behavioral Variability In LLMs

arXiv:2604.22027v1 Announce Type: cross Abstract: One of the most common complaints about large language models (LLMs) is their prompt sensitivity -- that is, the fact that their ability to perform a

applicationsarxiv-cs-ai

Applications

Sharpness-Aware Poisoning: Enhancing Transferability of Injective Attacks on Recommender Systems

arXiv:2604.22170v1 Announce Type: new Abstract: Recommender Systems~(RS) have been shown to be vulnerable to injective attacks, where attackers inject limited fake user profiles to promote the exposur

applicationsarxiv-cs-lg

.@shaunmmaguire says Elon Musk’s X acquisition was in service of civilizational-scale missions. 'I truly believe, like in my soul, that he i…

.@shaunmmaguire says Elon Musk’s X acquisition was in service of civilizational-scale missions. 'I truly believe, like in my soul, that he is trying to help humanity.' 'He has these goals. Like back u

industryelon-musk--x

Hardware

SIE3D: Single-Image Expressive 3D Avatar Generation via Semantic Embedding and Perceptual Expression Loss

arXiv:2509.24004v2 Announce Type: replace Abstract: Generating high-fidelity 3D head avatars from a single image is challenging, as current methods lack fine-grained, intuitive control over expression

hardwarearxiv-cs-cv

Local Ai

SNGR: Selective Non-Gaussian Refinement for Ambiguous SLAM Factor Graphs

arXiv:2604.22065v1 Announce Type: new Abstract: We present Selective Non-Gaussian Refinement (SNGR), a SLAM framework that augments iSAM2 with targeted nested sampling on windows where Gaussian approx

local-aiarxiv-cs-ro

so fun to see the reception to 5.5! there is almost nothing that feels more gratifying to me than builders saying they find our tools useful…

Sam Altman expresses satisfaction with the positive reception of OpenAI's 5.5 release, highlighting that developer feedback and builders finding value in OpenAI's tools is particularly rewarding to hi

industrysam-altman--x

SOC-ICNN: From Polyhedral to Conic Geometry for Learning Convex Surrogate Functions

arXiv:2604.22355v1 Announce Type: new Abstract: Classical ReLU-based Input Convex Neural Networks (ICNNs) are equivalent to the optimal value functions of Linear Programming (LP). This intrinsic struc

researcharxiv-cs-lg

Hardware

Soft Anisotropic Diagrams for Differentiable Image Representation

arXiv:2604.21984v1 Announce Type: new Abstract: We introduce Soft Anisotropic Diagrams (SAD), an explicit and differentiable image representation parameterized by a set of adaptive sites in the image

hardwarearxiv-cs-cv

Agents

SOLAR-RL: Semi-Online Long-horizon Assignment Reinforcement Learning

arXiv:2604.22558v1 Announce Type: cross Abstract: As Multimodal Large Language Models (MLLMs) mature, GUI agents are evolving from static interactions to complex navigation. While Reinforcement Learni

agentsarxiv-cs-ai

Agents

Sound Agentic Science Requires Adversarial Experiments

arXiv:2604.22080v1 Announce Type: new Abstract: LLM-based agents are rapidly being adopted for scientific data analysis, automating tasks once limited by human time and expertise. This capability is o

agentsarxiv-cs-ai

Agents

Source-Modality Monitoring in Vision-Language Models

arXiv:2604.22038v1 Announce Type: new Abstract: We define and investigate source-modality monitoring -- the ability of multimodal models to track and communicate the input source from which pieces of

agentsarxiv-cs-cl

Sources detail Microsoft's 'Windows K2', an ongoing initiative to address major Windows 11 user complaints about AI features, OS bloat, performance, and more (Zac Bowden/Windows Central)

Zac Bowden / Windows Central: Sources detail Microsoft's “Windows K2”, an ongoing initiative to address major Windows 11 user complaints about AI features, OS bloat, performance, and more — Microsoft

Sources: some UK officials fear Keir Starmer's plan for closer EU ties risks the US-UK alliance, and worry the UK may be forced to adopt the EU's AI regulation (Financial Times)

Financial Times: Sources: some UK officials fear Keir Starmer's plan for closer EU ties risks the US-UK alliance, and worry the UK may be forced to adopt the EU's AI regulation — Officials concerned a

Sovereign Agentic Loops: Decoupling AI Reasoning from Execution in Real-World Systems

model-releasesarxiv-cs-lg

arXiv:2604.22136v1 Announce Type: cross Abstract: Large language model (LLM) agents increasingly issue API calls that mutate real systems, yet many current architectures pass stochastic model outputs

SpaMEM: Benchmarking Dynamic Spatial Reasoning via Perception-Memory Integration in Embodied Environments

arXiv:2604.22409v1 Announce Type: new Abstract: Multimodal large language models (MLLMs) have advanced static visual--spatial reasoning, yet they often fail to preserve long-horizon spatial coherence

Tools

Speech translation in Google Meet is now rolling out to mobile devices

Speech translation in Google Meet is now rolling out to mobile devices I just encountered this feature via a 'try this out now' prompt in a Google Meet meeting. It kind-of worked! This is Google's imp

toolssimon-willison

Spend Less, Fit Better: Budget-Efficient Scaling Law Fitting via Active Experiment Selection

model-releasesarxiv-cs-lg

arXiv:2604.22753v1 Announce Type: new Abstract: Scaling laws are used to plan multi-million-dollar training runs, but fitting those laws can itself cost millions. In modern large-scale workflows, asse

Hardware

SpikingBrain2.0: Brain-Inspired Foundation Models for Efficient Long-Context and Cross-Platform Inference

arXiv:2604.22575v1 Announce Type: new Abstract: Scaling context length is reshaping large-model development, yet full-attention Transformers suffer from prohibitive computation and inference bottlenec

hardwarearxiv-cs-lg

Spontaneous Persuasion: An Audit of Model Persuasiveness in Everyday Conversations

arXiv:2604.22109v1 Announce Type: cross Abstract: Large language models (LLMs) possess strong persuasive capabilities that outperform humans in head-to-head comparisons. Users report consulting LLMs t

Spotify partners with Peloton to offer Premium subscribers access to curated playlists and 1,400+ ad-free fitness classes, its first foray into fitness content (Bloomberg)

Bloomberg: Spotify partners with Peloton to offer Premium subscribers access to curated playlists and 1,400+ ad-free fitness classes, its first foray into fitness content — Spotify Technology SA is no

SS3D: End2End Self-Supervised 3D from Web Videos

arXiv:2604.22686v1 Announce Type: new Abstract: We present SS3D, a web-scale SfM-based self-supervision pretraining pipeline for feed-forward 3D estimation from monocular video. Our model jointly pred

researcharxiv-cs-cv

SSG: Logit-Balanced Vocabulary Partitioning for LLM Watermarking

arXiv:2604.22438v1 Announce Type: cross Abstract: Watermarking has emerged as a promising technique for tracing the authorship of content generated by large language models (LLMs). Among existing appr

StateX: Enhancing RNN Recall via Post-training State Expansion

arXiv:2509.22630v3 Announce Type: replace-cross Abstract: Recurrent neural networks (RNNs), such as linear attention and state-space models, have gained popularity due to their constant per-token comp

STEM: Structure-Tracing Evidence Mining for Knowledge Graphs-Driven Retrieval-Augmented Generation

arXiv:2604.22282v1 Announce Type: new Abstract: Knowledge Graph-based Question Answering (KGQA) plays a pivotal role in complex reasoning tasks but remains constrained by two persistent challenges: th

researcharxiv-cs-cl