All entries

I've been building a C99 inference engine from scratch (no Python, no BLAS, just gcc and make) that runs BitNet's ternary models on CPU. A few weeks ago I got obsessed with the matmul kernel - wrote a

local-air-localllama

Hardware

SPORD: A Simulation-Propose-then-OR-Dispose Approach for Supply Chain Planning

DGX agent

arXiv:2607.21354v1 Announce Type: new Abstract: For years, supply chain planning at e-commerce firms has operated as a collection of isolated projects. Each planning task from static network planning

hardwarearxiv-cs-ai

24 Jul 2026

SR-TTT Does Not Learn Retrieval: A Correction and Mechanistic Post-Mortem of Surprisal-Aware Residual Test-Time Training

arXiv:2603.06642v2 Announce Type: replace-cross Abstract: Test-Time Training (TTT) language models replace the KV-cache with fast weights updated during inference, achieving O(1) memory but suffering

tutorialsarxiv-cs-ai

StabilityBench: Benchmarking Instability in LLMs

arXiv:2607.20558v1 Announce Type: cross Abstract: AI Assistants are increasingly deployed in high-stakes settings, such as healthcare or government services. Yet their real-world behavior remains poor

Statistical Inference for Generative Model Comparison

model-releasesarxiv-cs-lg

arXiv:2501.18897v4 Announce Type: replace-cross Abstract: Generative models have achieved remarkable success across a range of applications, yet their evaluation still lacks principled uncertainty qua

Statistical physics of deep learning: Optimal learning of a multi-layer perceptron near interpolation

arXiv:2510.24616v4 Announce Type: replace-cross Abstract: For four decades statistical physics has been providing a framework to analyse neural networks. A long-standing question remained on its capac

tutorialsarxiv-cs-lg

Statistics of week: “More Americans pay for sports betting apps (5%) than pay for AI… 37% of consumers say none of AI's uses are helpful. … …

Statistics of week: “More Americans pay for sports betting apps (5%) than pay for AI… 37% of consumers say none of AI's uses are helpful. … 2.2% penetration nearly four years after ChatGPT launched. N

safetygary-marcus--x

Stay hungry, stay foolish. Absolute legend

Stay hungry, stay foolish. Absolute legend For my first post, I’m sharing a letter @NVIDIA signed on why open models matter. AI will transform every industry, power every company, and be built by ever

safetyjim-fan--x

STeMP: Spatio-Temporal Modelling Protocol

arXiv:2607.20592v1 Announce Type: new Abstract: Spatio-temporal machine-learning modelling is an important tool in environmental research. However, machine-learning models are highly sensitive to both

tutorialsarxiv-cs-lg

Stochastic Sampling is Epistemically Shallow: The Dimensionality Gap Between Temperature Variation and Model Diversity in LLMs

arXiv:2607.20464v1 Announce Type: new Abstract: When a language model gives different answers on repeated runs, does that variation reveal what it does not know? Self-consistency turns the variation i

Stokes-Informed Diffusion for Robust Linear Polarization Estimation

arXiv:2607.21239v1 Announce Type: new Abstract: Polarization cues benefit applications such as material detection and de-reflection, yet acquiring them typically requires dedicated hardware. This moti

safetyarxiv-cs-cv

Agents

Streaming Multi-Agent Autoregressive Diffusion Model with World State Registers

arXiv:2607.21594v1 Announce Type: new Abstract: Multi-agent interactive world models should not only generate consistent observations, but also maintain world states that persist across agents and evo

agentsarxiv-cs-cv

StrideDiffusion: Accelerating Diffusion Models for Time-series Generation

arXiv:2607.20545v1 Announce Type: new Abstract: Diffusion models have become competitive generators for time series, but their practical use is limited by the large number of sequential denoising step

Structure-Preserving Physics-Informed Neural Network for the Korteweg--de Vries (KdV) Equation

arXiv:2511.00418v2 Announce Type: replace Abstract: Physics-Informed Neural Networks (PINNs) offer a flexible framework for solving nonlinear partial differential equations (PDEs), yet conventional im

Student-Centered Distillation Narrows the Agentic Gap Between Small and Large LLMs

arXiv:2509.14257v3 Announce Type: replace-cross Abstract: Large Language Model agents achieve strong performance on multi-step reasoning and tool-use tasks, but their impressive capabilities typically

SubSplat: High-Resolution Pixel-aligned 3DGS via Sub-pixel Gaussian Reparameterization

arXiv:2607.20813v1 Announce Type: new Abstract: Pixel-aligned Gaussian splatting enables efficient and generalizable novel-view synthesis. However, high-resolution rendering faces a critical trade-off

researcharxiv-cs-cv

SuperFlow: Training Flow Matching Models with RL on the Fly

arXiv:2512.17951v3 Announce Type: replace Abstract: Recent progress in flow-based generative models and reinforcement learning (RL) has improved text-image alignment and visual quality. However, curre

safetyarxiv-cs-cv

Surprisal Theory is Tautological (without Rational Grounding)

arXiv:2607.21574v1 Announce Type: new Abstract: Surprisal theory holds that the human processing difficulty of a linguistic unit in context is an affine function of its surprisal under some language m

researcharxiv-cs-cl

swiss-ai/Apertus-v1.5 70B/8B

model-releasesr-localllama

https://huggingface.co/swiss-ai/Apertus-v1.5-70B https://huggingface.co/swiss-ai/Apertus-v1.5-8B Apertus 1.5 is a family of 8B and 70B parameter language models designed to advance the state of multil

Hardware

Synopsys targets physical AI complexity with co-design and agentic chip workflows

The rapid evolution of intelligent software-defined systems is pushing chip design into a new era of physical AI, one where chip design complexity is outpacing traditional engineering methods. Manufac

hardwaresiliconangle

Applications

Synthetic data generation framework for quality control automation in gravure printing

arXiv:2607.21577v1 Announce Type: cross Abstract: Quality control in printing, particularly in rotogravure printing, still depends on slow, costly, and subjective manual inspection. Automated surface

applicationsarxiv-cs-ai

Synthetic minority data is redundant or invalid: a data-dependent validity theory and a de-biased test

arXiv:2607.20787v1 Announce Type: cross Abstract: For two decades, the standard remedy for class-imbalanced learning has been to fabricate synthetic minority examples, and the standard evidence of the

safetyarxiv-cs-ai

T-STAR: A Large-Scale Benchmark for Spatio-Temporal Panoptic Scene Graph Generation in Satellite Video

model-releasesarxiv-cs-cv

arXiv:2607.21228v1 Announce Type: new Abstract: Structured understanding of satellite video is essential for advancing dynamic geospatial scene analysis from low-level perception to high-level cogniti

Applications

TableVerse: A Large-scale Tabletop Dataset with Real-world Grounded Layouts for Generalizable Manipulation

arXiv:2607.21017v1 Announce Type: new Abstract: The development of generalizable robotic manipulation policies is inherently bounded by the availability of large-scale, high-fidelity scene data. While

applicationsarxiv-cs-ro

TabPFN beyond Tabular Data: Calibration and Accuracy on Multimodal Embeddings

arXiv:2607.11007v2 Announce Type: replace Abstract: Few-shot multimodal classification commonly attaches a lightweight head, such as k-nearest neighbors, logistic regression, or a linear SVM, to a fro

Tackling Heterogeneity in Federated Learning via Variance-Reduced Boltzmann Sampling within Homogeneous Social Coalitions

arXiv:2506.02897v3 Announce Type: replace Abstract: Federated Learning (FL) enables privacy-preserving collaborative model training, but its effectiveness is often limited by client data heterogeneity

Agents

TeaRAG: A Token-Efficient Agentic Retrieval-Augmented Generation Framework

arXiv:2511.05385v2 Announce Type: replace-cross Abstract: Retrieval-Augmented Generation (RAG) utilizes external knowledge to augment Large Language Models' (LLMs) reliability. For flexibility, agenti

agentsarxiv-cs-ai

Telco-GAIA: Bilingual Benchmark for Agents in Telecom Domain

arXiv:2607.20510v1 Announce Type: new Abstract: We introduce Telco-GAIA, a bilingual, multi-modal benchmark for evaluating tool-using agents on the data of a real-world telecommunications operator. Te

Tencent WorkBuddy Bench: A Multi-Domain Coding-Agent Benchmark with Contamination-Resistant Task Construction

model-releasesarxiv-cs-cl

arXiv:2607.20911v1 Announce Type: new Abstract: We introduce Tencent WorkBuddy Bench, a multi-domain evaluation suite for coding agents; this report documents its construction methodology, scoring pro

Local Ai

Test-Time Scaling via Error Localization

arXiv:2607.21453v1 Announce Type: new Abstract: Scaling inference-time computation has emerged as a reliable method to improve the performance of large language models on complex reasoning and program

local-aiarxiv-cs-lg

Texture++: Elevating 3D Asset Texture Resolution with a Region-Aware Diffusion Model

arXiv:2607.21504v1 Announce Type: new Abstract: Numerous 3D assets are discarded due to low texture resolution, while current super-resolution models ignore texture maps and focus on natural images. A

researcharxiv-cs-cv

thaulab@EEUCA 2026: Who Said What to Whom? A Targeting-Aware Neural-Symbolic Pipeline for Gaming Toxicity Detection

arXiv:2607.20447v1 Announce Type: new Abstract: This paper describes our system for the EEUCA 2026 Shared Task on toxicity classification in gaming chat. We implement a three-stage pipeline combining

safetyarxiv-cs-cl

The Active Ingredient in Muon's Grokking

arXiv:2607.20512v1 Announce Type: cross Abstract: The Muon optimizer reaches the grokking threshold on modular arithmetic faster than AdamW. Prior work attributes this to 'spectral-norm constraints pl

Applications

The Boundaries of Automation: A Theory of Persistent Human Participation

arXiv:2607.21547v1 Announce Type: new Abstract: The rapid progress of AI has intensified the long-standing pursuit of automation: replacing human participation with algorithms wherever possible. Impli

applicationsarxiv-cs-ai

The Dark Room in the Reward Channel: Dense Prediction Rewards Collapse GRPO-Trained LLM Agents -- and What Actually Works

arXiv:2607.21273v1 Announce Type: new Abstract: Dense per-step supervision is an appealing remedy for sparse-reward, long-horizon LLM agents: reward the agent for predicting its next observation, and

safetyarxiv-cs-lg

Local Ai

The demand for GLM-5.2 and other frontier-level open models has been surging on Ollama's cloud. We're adding capacity in anticipation for so…

The demand for GLM-5.2 and other frontier-level open models has been surging on Ollama's cloud. We're adding capacity in anticipation for some very large models next week! To make sure Ollama's infras

local-aiollama--x

The Devil is in the Spectrum: Mitigating Representation Collapse in LLMs via Topologically Regularized Side-Path

arXiv:2607.20484v1 Announce Type: new Abstract: Large Language Models (LLMs) are fundamentally limited by representation collapse, a bottleneck that severely degrades long-context performance. We iden

The 'distillation' claim is just ridiculous in nature

Even if China was distilling from US models (assuming all accusations are true), nothing about it makes it illegal. It is like saying you distilled knowledge from your professor in colleges and now he

tutorialsr-localllama

The Geometry of Personality: Activation Steering with Jungian Cognitive Functions

arXiv:2607.20803v1 Announce Type: cross Abstract: Activation steering enables control and interpretation of LLMs, yet existing work primarily models personality through static trait frameworks such as

The GPT-5.x Pro series has remained the best models for hard technical problems since they launched. There is some parallel model magic goin…

model-releasesethan-mollick--x

The GPT-5.x Pro series has remained the best models for hard technical problems since they launched. There is some parallel model magic going on that is not well-explained. Anthropic has never had an

The Hidden Footprint: Making Storage a First-Class Metric for LLM Agent Evaluation

arXiv:2607.11149v3 Announce Type: replace Abstract: LLM agent benchmarks measure task completion, reliability, and inference cost, but not the persistent data an agent run leaves on disk, including lo

The Human-AI Substitution Principle: When will you be replaced by AI in your organization?

arXiv:2607.20781v1 Announce Type: new Abstract: Artificial Intelligence (AI) is rapidly transforming organizations, raising a fundamental organizational and economic question: when will a human employ

The knowledge that makes AI useful is diffused. It lives with scientists, engineers, clinicians, firms. For AI to benefit from distributed k…

The knowledge that makes AI useful is diffused. It lives with scientists, engineers, clinicians, firms. For AI to benefit from distributed knowledge, it must itself be distributed. Agree with Jensen t

safetyelon-musk--x

The Price of Hidden Curvature: An widetilde{Omega} (d^{5/4} sqrt{T}) Lower Bound for Bandit Convex Optimization

arXiv:2607.18652v2 Announce Type: replace-cross Abstract: We establish a widetildeOmega(d^{5/4}sqrt T) lower bound on the minimax expected regret of stochastic bandit convex optimization of 1-Lipschit

The RealDefocus Benchmark for Defocus Deblurring

model-releasesarxiv-cs-cv

arXiv:2607.21078v1 Announce Type: new Abstract: Single-Image Defocus Deblurring (SIDD) aims to recover an all-in-focus image from a single defocused observation, but rigorous and reproducible evaluati

The Second LoViF 2026 Challenge on Real-World All-in-One Image Restoration: Methods and Results

model-releasesarxiv-cs-cv

arXiv:2607.21118v1 Announce Type: new Abstract: This paper presents a review of the second LoViF Challenge on Real-World All-in-One Image Restoration. The challenge aims to advance unified image resto

The Storyteller in the Model: Narrative Pattern Inheritance, Escalation Dynamics, and Alignment Governance in LLMs

arXiv:2607.20449v1 Announce Type: cross Abstract: LLMs are trained predominantly on human-authored text, yet the structural and narrative conventions embedded in that text are rarely examined as a sou

safetyarxiv-cs-ai

The Washington Post processed 1.79B input tokens per month through Together AI, running open models like Llama and Mistral in production wit…

model-releasestogether-ai--x

The Washington Post processed 1.79B input tokens per month through Together AI, running open models like Llama and Mistral in production with predictable costs and full control over the model stack. T