All entries

Self-CriTeach: LLM Self-Teaching and Self-Critiquing for Improving Robotic Planning via Automated Domain Generation

arXiv:2509.21543v3 Announce Type: replace Abstract: Large Language Models (LLMs) have shown strong promise for robotic task planning, particularly through the automatic generation of symbolic planning

researcharxiv-cs-ro

Safety

Self-Supervised On-Policy Reinforcement Learning via Contrastive Proximal Policy Optimisation

DGX agent

arXiv:2605.13554v1 Announce Type: cross Abstract: Contrastive reinforcement learning (CRL) learns goal-conditioned Q-values through a contrastive objective over state-action and goal representations,

safetyarxiv-cs-ai

14 May 2026

Semantic knowledge guides innovation and drives cultural evolution

arXiv:2510.12837v3 Announce Type: replace-cross Abstract: Cultural evolution allows ideas and technologies to accumulate across generations, reaching their most complex and open-ended form in humans.

agentsarxiv-cs-ai

Senses Wide Shut: A Representation-Action Gap in Omnimodal LLMs

model-releasesarxiv-cs-ai

arXiv:2605.13737v1 Announce Type: new Abstract: When an omnimodal large language model accepts a question whose textual premise contradicts what it actually sees or hears, does the failure lie in perc

Separating Shortcut Transition from Cross-Family OOD Failure in a Minimal Model

arXiv:2605.12945v1 Announce Type: new Abstract: Shortcut features are often invoked to explain out-of-distribution (OOD) failure, but training correlation, learned shortcut use, and test-time failure

Applications

Seven papers. One research team. Together AI is heading to #MLSys2026 next week. Check out the work going from research to production on the…

applicationstogether-ai--x

Together AI will present seven research papers at MLSys 2026, showcasing projects that demonstrate the transition from research to production applications. The announcement highlights the company's co

Sharpness-Guided Group Relative Policy Optimization via Probability Shaping

arXiv:2511.00066v4 Announce Type: replace Abstract: Reinforcement learning with verifiable rewards (RLVR) has become a practical route to improve large language model reasoning, and Group Relative Pol

safetyarxiv-cs-lg

Ship it. Measure it. All in one flow. @Mixpanel is on Replit MCP. See it live at the London hackathon next week 👇

Ship it. Measure it. All in one flow. @Mixpanel is on Replit MCP. See it live at the London hackathon next week 👇 Mixpanel is live in the Replit MCP marketplace! Devs building on @Replit can pull prod

agentsreplit--x

SHM-Agents: A Generalist-Specialist Integrated Agent System for Structural Health Monitoring

arXiv:2605.12916v1 Announce Type: cross Abstract: Artificial intelligence is increasingly used to simplify complex tasks. In engineering applications of structural health monitoring (SHM), existing sp

agentsarxiv-cs-lg

Tutorials

Shortcut Mitigation via Spurious-Positive Samples

arXiv:2605.13340v1 Announce Type: new Abstract: Shortcut mitigation strategies commonly rely on training data annotations, group-balanced held-out data or the presence of all groups, i.e., all combina

tutorialsarxiv-cs-lg

SID: Sliding into Distribution for Robust Few-Demonstration Manipulation

arXiv:2605.13428v1 Announce Type: new Abstract: Generalizing robotic manipulation across object poses, viewpoints, and dynamic disturbances is difficult, especially with only a few demonstrations. End

safetyarxiv-cs-ro

Simulating Students or Sycophantic Problem Solving? On Misconception Faithfulness of LLM Simulators

arXiv:2605.12748v1 Announce Type: cross Abstract: Large language models (LLMs) can fluently generate student-like responses, making them attractive as simulated students for training and evaluating AI

researcharxiv-cs-ai

Hardware

Since going on @creatine_cycle podcast 1.5 months ago, I have lost 4.8 pounds of fat and gained 4.6 pounds of muscle. Swole as a service wor…

Since going on @creatine_cycle podcast 1.5 months ago, I have lost 4.8 pounds of fat and gained 4.6 pounds of muscle. Swole as a service works. on today's episode of Swole as a Service i brought my go

hardwaredylan-patel--x

sketch2symm: Symmetry-aware sketch-to-shape generation via semantic bridging

arXiv:2510.11303v2 Announce Type: replace Abstract: Sketch-based 3D reconstruction remains a challenging task due to the abstract and sparse nature of sketch inputs, which often lack sufficient semant

researcharxiv-cs-cv

Skill-Aligned Annotation for Reliable Evaluation in Text-to-Image Generation

arXiv:2605.13223v1 Announce Type: new Abstract: Text-to-image (T2I) generation has advanced rapidly, making reliable evaluation critical as performance differences between models narrow. Existing eval

researcharxiv-cs-cv

SkySplat: Generalizable 3D Gaussian Splatting from Multi-Temporal Sparse Satellite Images

model-releasesarxiv-cs-cv

arXiv:2508.09479v2 Announce Type: replace Abstract: Three-dimensional scene reconstruction from sparse-view satellite images is a long-standing and challenging task. While 3D Gaussian Splatting (3DGS)

SMA: Submodular Modality Aligner For Data Efficient Multimodal Learning

arXiv:2605.12872v1 Announce Type: new Abstract: Despite the recent success of Multimodal Foundation Models (FMs), their reliance on massive paired datasets limits their applicability in low-data and r

Small Area Estimation of Case Growths for Timely COVID-19 Outbreak Detection

arXiv:2312.04110v2 Announce Type: replace-cross Abstract: The COVID-19 pandemic has exerted a profound impact on the global economy and continues to exact a significant toll on human lives. The COVID-

SmithDB is a feat of engineering. A new database for a new data shape Huge work from @ankush_gola11 and team

SmithDB is a feat of engineering. A new database for a new data shape Huge work from @ankush_gola11 and team We built SmithDB: the database purpose built for agent observability workloads that now pow

agentsharrison-chase--x

SmithDB is the perfect example of how far performance can be pushed by having full control over the storage layer. The DataFusion + @vortexd…

SmithDB is the perfect example of how far performance can be pushed by having full control over the storage layer. The DataFusion + @vortexdotdev stack seems to be emerging as THE way to build next ge

agentsharrison-chase--x

so cool. win for the shape rotators! @ericho_goodfire

so cool. win for the shape rotators! @ericho_goodfire Neural networks do math by rotating shapes. We found a shape-rotating calculator hidden inside an LLM – and it’s used for more than just math! (1/

industrysonya-huang--x

Sockpuppetting: Jailbreaking LLMs by Combining Prefilling with Optimization

arXiv:2601.13359v2 Announce Type: replace-cross Abstract: Prefill attacks are an effective and low-cost jailbreaking method, as they directly insert an acceptance sequence (e.g., 'Sure, here is how to

SoK: A Comprehensive Analysis of the Current Status of Neural Tangent Generalization Attacks with Research Directions

arXiv:2605.12792v1 Announce Type: new Abstract: There is recently a serious issue that Deep Neural Networks (DNNs) training uses more and more unauthorized data. A clean-label generalization attack, o

Local Ai

Someone posted a real Monet to twitter but said it was AI generated. The replies are amazing, pretentious and confidently wrong

local-air-stablediffusion

A user posted an actual Claude Monet Water Lilies painting while claiming it was AI-generated and asking critics to explain its inferiority to real Monet work. The detailed critiques were specific and

Sources: 50+ researchers and engineers have left xAI since the SpaceX acquisition via layoffs, firings, and voluntary departures; many have joined Meta and TML (Theo Wayt/The Information)

Theo Wayt / The Information: Sources: 50+ researchers and engineers have left xAI since the SpaceX acquisition via layoffs, firings, and voluntary departures; many have joined Meta and TML — Call it t

Sources: Anthropic has agreed to the terms of a 30B fundraising at a 900B valuation, with Sequoia, Dragoneer, Greenoaks, and Altimeter co-leading the round (George Hammond/Financial Times)

George Hammond / Financial Times: Sources: Anthropic has agreed to the terms of a 30B fundraising at a 900B valuation, with Sequoia, Dragoneer, Greenoaks, and Altimeter co-leading the round — AI lab h

Sources: climate watchdog SBTi drops proposed rules that would limit tech companies' claims that gas-fueled data centers are covered by clean energy investments (Kenza Bryan/Financial Times)

Kenza Bryan / Financial Times: Sources: climate watchdog SBTi drops proposed rules that would limit tech companies' claims that gas-fueled data centers are covered by clean energy investments — Corpor

Applications

Sources: Intel has begun testing production of 'low-end/legacy iPhone, iPad, and Mac processors'; Apple thinks TSMC's resources will continue tilting toward AI (@mingchikuo)

@mingchikuo: Sources: Intel has begun testing production of “low-end/legacy iPhone, iPad, and Mac processors”; Apple thinks TSMC's resources will continue tilting toward AI — ... Apple has kicked off

applicationstechmeme

Sources: Joshua Kushner's Thrive Capital told its stakeholders that it has invested ~$100M in Shopify, framed as a bet on how AI could lead to gains in commerce (Natasha Mascarenhas/Bloomberg)

Natasha Mascarenhas / Bloomberg: Sources: Joshua Kushner's Thrive Capital told its stakeholders that it has invested ~$100M in Shopify, framed as a bet on how AI could lead to gains in commerce — Josh

Sources: Microsoft plans to remove most of its Claude Code licenses and push its developers toward GitHub Copilot CLI, after previously pushing Claude Code (Tom Warren/The Verge)

Tom Warren / The Verge: Sources: Microsoft plans to remove most of its Claude Code licenses and push its developers toward GitHub Copilot CLI, after previously pushing Claude Code — Thousands of Micro

model-releasestechmeme

Sources: xAI co-founder Igor Babuschkin plans to raise up to 1B at an up to 5B valuation for a new AI research startup, with General Catalyst possibly leading (Forbes)

Forbes: Sources: xAI co-founder Igor Babuschkin plans to raise up to 1B at an up to 5B valuation for a new AI research startup, with General Catalyst possibly leading — He's the latest superstar AI re

SP-GCRL: Influence Maximization on Incomplete Social Graphs

arXiv:2605.12513v1 Announce Type: cross Abstract: Influence maximization (IM) in real platforms is challenged by incomplete, noisy social graphs and non-stationary diffusion dynamics. We propose SP-GC

safetyarxiv-cs-ai

Tutorials

Sparse Code Uplifting for Efficient 3D Language Gaussian Splatting

arXiv:2605.13600v1 Announce Type: new Abstract: 3D Language Gaussian Splatting (3DLGS) augments 3D Gaussian Splatting with language-aligned visual features for open-vocabulary 3D scene understanding.

tutorialsarxiv-cs-cv

SpatialReward: Bridging the Perception Gap in Online RL for Image Editing via Explicit Spatial Reasoning

model-releasesarxiv-cs-cv

arXiv:2602.07458v4 Announce Type: replace Abstract: Online Reinforcement Learning (RL) offers a promising avenue for complex image editing but is currently constrained by the scarcity of reliable and

Spatiotemporal downscaling and nowcasting of urban land surface temperatures with deep neural networks

arXiv:2605.13566v1 Announce Type: new Abstract: Land Surface Temperature (LST) is a key variable for various applications, such as urban climate and ecology studies. Yet, existing satellite-derived LS

safetyarxiv-cs-lg

Spectral Energy Centroid: a Metric for Improving Performance and Analyzing Spectral Bias in Implicit Neural Representations

arXiv:2605.12709v1 Announce Type: new Abstract: Implicit Neural Representations (INRs) model continuous signals using multilayer perceptrons (MLPs), enabling compact, differentiable, and high-fidelity

safetyarxiv-cs-lg

Spectral Flattening Is All Muon Needs: How Orthogonalization Controls Learning Rate and Convergence

arXiv:2605.13079v1 Announce Type: cross Abstract: Muon orthogonalizes the momentum buffer before each update, replacing its singular values with ones via Newton-Schulz iterations. This simple change l

researcharxiv-cs-ai

SpectralTrain: A Universal Framework for Hyperspectral Image Classification

model-releasesarxiv-cs-ai

arXiv:2511.16084v2 Announce Type: replace-cross Abstract: Hyperspectral image (HSI) classification typically involves large-scale data and computationally intensive training, which limits the practica

SpikeProphecy: A Large-Scale Benchmark for Autoregressive Neural Population Forecasting

arXiv:2605.12992v1 Announce Type: cross Abstract: Neural population models, which predict the joint firing of many simultaneously recorded neurons forward in time, are typically evaluated by a single

SPOT: Selective Prompt Projection via Total Variation for Inference-Only Safe Text-to-Image Generation

arXiv:2602.00616v3 Announce Type: replace Abstract: Text-to-Image (T2I) diffusion models enable high quality open ended synthesis, but practical use requires suppressing unsafe generations while prese

safetyarxiv-cs-ai

SpurAudio: A Benchmark for Studying Shortcut Learning in Few-Shot Audio Classification

model-releasesarxiv-cs-cv

arXiv:2605.13672v1 Announce Type: new Abstract: Few-shot classification (FSC) is widely used for learning from limited labeled data, yet most evaluations implicitly assume that target concepts are ind

Applications

SSDA: Bridging Spectral and Structural Gaps via Dual Adaptation for Vision-Based Time Series Forecasting

arXiv:2605.12550v1 Announce Type: cross Abstract: Large vision models (LVMs) have recently proven to be surprisingly effective time series forecasters, simply by rendering temporal data as images. Thi

applicationsarxiv-cs-ai

Stable Attention Response for Reliable Precipitation Nowcasting

arXiv:2605.13181v1 Announce Type: cross Abstract: Precipitation nowcasting remains challenging due to the highly localized, rapidly evolving, and heterogeneous nature of atmospheric dynamics. Although

researcharxiv-cs-ai

STAR: Semantic-Temporal Adaptive Representation Learning for Few-Shot Action Recognition

arXiv:2605.13202v1 Announce Type: cross Abstract: Few-shot action recognition (FSAR) requires models to generalize to novel action categories from only a handful of annotated samples. Despite progress

safetyarxiv-cs-ai

State-Centric Decision Process

arXiv:2605.12755v1 Announce Type: new Abstract: Language environments such as web browsers, code terminals, and interactive simulations emit raw text rather than states, and provide none of the runtim

agentsarxiv-cs-ai

State-of-art minibatches via novel DPP kernels: discretization, wavelets, and rough objectives

arXiv:2605.13127v1 Announce Type: cross Abstract: Determinantal point processes (DPPs) have emerged as a kernelized alternative to vanilla independent sampling for generating efficient minibatches, co

State-Space NTK Collapse Near Bifurcations

arXiv:2605.12763v1 Announce Type: new Abstract: Rich feature learning in tasks that unfold over time often requires the model to pass through bifurcations, constituting qualitative changes in the unde

Steer-to-Detect: Probing Hidden Representations for Detection of LLM-Generated Texts

arXiv:2605.12890v1 Announce Type: cross Abstract: The rapid advancement of large language models (LLMs) has made machine-generated text increasingly difficult to distinguish from human-written text. W