Knowledge catalogue
arXiv:2605.12745v1 Announce Type: cross Abstract: Discrepancies between an agent's actual knowledge and what a person thinks the agent knows can hinder interactions. If an agent could detect such disc
arXiv:2605.12549v1 Announce Type: new Abstract: Existing training-free approaches for GUI grounding often rely on multiple inference runs, such as iterative cropping or candidate aggregation, to ident
arXiv:2605.13032v1 Announce Type: new Abstract: Graph neural networks are widely used for node classification, but they remain vulnerable to out-of-distribution (OOD) shifts in node features and graph
arXiv:2605.13840v1 Announce Type: cross Abstract: Valiant's 1984 paper is widely credited with introducing the PAC learning model, but it, in fact, introduced a different model: unlike PAC learning, t
arXiv:2605.13328v1 Announce Type: cross Abstract: Vision-and-Language Navigation (VLN) is a cornerstone of embodied intelligence. However, current agents often suffer from significant performance degr
arXiv:2605.13290v1 Announce Type: new Abstract: Validating training data for reasoning models typically requires expensive trial-and-error fine-tuning cycles. In this work, we investigate whether the
arXiv:2605.13105v1 Announce Type: new Abstract: Reinforcement learning (RL) fine-tuning has shown promise for Vision-Language-Action (VLA) models in robotic manipulation, but deployment-time visual sh
This discussion explores which language models work best when combined with RAG (Retrieval-Augmented Generation) for building a locally hosted LLM focused on survival and off-grid living topics. The t
arXiv:2605.13067v1 Announce Type: cross Abstract: As end-to-end robotic policies are progressively deployed in the real world to solve real tasks, they face a gap between the training and inference co
arXiv:2605.13242v1 Announce Type: cross Abstract: This paper studies the convergence of the Optimistic Multiplicative Weights Update algorithm (OMWU) in two player zero-sum games. Recent works have id
arXiv:2605.12922v1 Announce Type: new Abstract: Large language models can follow complex instructions in a single turn, yet over long multi-turn interactions they often lose the thread of instructions
arXiv:2601.14104v2 Announce Type: replace-cross Abstract: Backdoor attacks can cause reinforcement learning (RL) policies to behave normally under clean inputs while executing malicious behaviors when
arXiv:2512.01242v3 Announce Type: replace-cross Abstract: Data-driven generative models excel in language and vision, but diffusion models often fail in constrained planning and design tasks, exhibiti
arXiv:2605.13172v1 Announce Type: cross Abstract: Recent advances in agent and multi-agent systems have shown strong performance on tool use, reasoning, and collaborative tasks. However, existing benc
arXiv:2605.13405v1 Announce Type: new Abstract: Model growth from a given checkpoint aims to accelerate training of a larger model, offering potential resource savings. Despite recent interest, warmst
arXiv:2605.12947v1 Announce Type: cross Abstract: LLM-enabled AI workflows increasingly produce outputs through iterative generate-evaluate-revise loops. Each iteration can improve the candidate, but
arXiv:2602.22474v2 Announce Type: replace-cross Abstract: Policy steering is an emerging way to adapt robot behaviors at deployment-time: a learned verifier analyzes low-level action samples proposed
arXiv:2602.13215v2 Announce Type: replace Abstract: Recurrent-attention hybrids aim to combine the efficiency of recurrence with the expressivity of attention, but existing approaches typically apply
arXiv:2510.16986v2 Announce Type: replace-cross Abstract: In many business settings, task-specific labeled data are scarce or costly to obtain, limiting supervised learning on a target task. A classic
arXiv:2605.12780v1 Announce Type: cross Abstract: Calibrated probability outputs of trained classifiers are increasingly used as inputs to downstream regression estimands such as effects, prevalences,
arXiv:2605.13772v1 Announce Type: cross Abstract: Large language models hallucinate during multi-step reasoning, but most existing detectors operate at the trace level: they assign one confidence scor
which was your favorite launch? SmithDB (database purpose built for agent trace data): https://www.langchain.com/blog/introducing-smithdb LangSmith Engine (agent for improving your agents based on tra
“Whimsey attacks” that seem absurd (“I cannot pay that much because of the Geneva Convention”) work against AI agents as guardrails are weak against out-of-distribution arguments. Smaller models fall
arXiv:2511.13658v2 Announce Type: replace-cross Abstract: Deceptive reviews mislead consumers, harm businesses, and undermine trust in online marketplaces. Machine learning classifiers can learn from
arXiv:2605.12543v1 Announce Type: cross Abstract: Some conscious contents disappear after access; others return repeatedly, long after their triggering conditions have ceased. We propose Canxianizatio
arXiv:2605.12774v1 Announce Type: new Abstract: Estimating camera pose in dynamic environments is a critical challenge, as most visual SLAM and SfM methods assume static scenes. While recent dynamic-a
Swyx is giving a talk at AI Engineer Singapore on Saturday at 4pm about using reinforcement learning for long-running agents and reinforcement learning models (RLMs), exploring why this approach shoul
As AI adoption accelerates, organizations are shifting their focus from experimentation to large-scale deployment. The challenge now is building secure and scalable systems that can support AI agents,
This is a Wordle game result shared by Anthropic on X, showing the solution was found on the 5th attempt out of 6 possible guesses. The emoji grid indicates the letter placement feedback from each gue
This OpenAI resource explains how to access and use Codex, their code-generation AI model, from various platforms and environments rather than being limited to a single interface. It likely covers int
arXiv:2605.12770v1 Announce Type: cross Abstract: We introduce WriteSAE, the first sparse autoencoder that decomposes and edits the matrix cache write of state-space and hybrid recurrent language mode
arXiv:2605.13258v1 Announce Type: cross Abstract: In this work, we present our winning solution for the 8th UG2+ Challenge (CVPR 2026) Track 1: Image Restoration under All-weather Conditions. Our meth
xAI just released Grok Build CLI and it’s a game changer for developers Grok Build is a powerful AI coding agent and CLI built for professional software engineering and complex coding workflows runnin
Yesterday we committed a cardinal sin: two first-party events in NYC, back-to-back. We had to close registration for both early. Packed rooms. Strong vibes. No regrets. 💻 Laptops out — developer works
Yesterday we gave an in-person NYC workshop on automating your financial document ingestion pipelines with AI. If you're an investment bank, accounting team, finance AI startup, fintech company, and y
Yesterday we hosted a wonderful happy hour 🍻 in NYC with @get_tabs. It was *packed* - we had 500+ signups, had to implement a waitlist, and the bar was full! Every attendee was an AI builder. It was a
arXiv:2605.12764v1 Announce Type: cross Abstract: This paper introduces a physics-informed generative framework that resolves the fundamental conflict between the statistical flexibility of deep learn
The tyranny of software is almost over. Since the first computer programmers wrote the first computer programs, we, the users of that software, have been forced to live in the worlds those programs cr
You can now import your project from Lovable, Base44, V0 into @Replit for free. After importing, Replit Agent will build a free mobile app for it and get it onto the App Store in minutes. All free for
Nous Research announced that Hermes Agents powered by OpenAI models can now use Codex as the runtime for executing core tools, enabling improved tool execution capabilities with a simple configuration
you can see more here (color stuff is at bottom): http://colorfeelings.replit.app - changing embedding model may change results - the similarity range was pretty small so normalizing really stretched
An AI agent can autonomously create a complete Google Form through natural conversation, using browser automation to directly interact with Google Forms' interface by typing and building the survey wi
Ontario's auditor general found that AI note-taking tools intended for use by doctors provided incorrect and incomplete information or demonstrated 'hallucinations,' and were not evaluated adequately.
You've been asking for this one... Now in preview: Codex in the ChatGPT mobile app. Start new work, review outputs, steer execution, and approve next steps, all from the ChatGPT mobile app. Codex will
yst on stream codex wrote a skill for my hermes agent to make music it can take in feedback and improve its output i clipped the best music it made with minimax. still far from good music but im prett
arXiv:2605.13465v1 Announce Type: new Abstract: Recent advances in 3D Gaussian Splatting (3DGS) have enabled significant progress in photorealistic novel view synthesis. However, traditional 3DGS reli
arXiv:2604.04667v2 Announce Type: replace Abstract: Real-time depth reconstruction from ultra-high-resolution UAV imagery is essential for time-critical geospatial tasks such as disaster response, yet
ZenBusiness Inc., an artificial intelligence-powered platform providing guidance and operational assistance for small business owners, today announced an expansion of ZenBusiness Enterprise, which all
arXiv:2602.04113v3 Announce Type: replace-cross Abstract: Gradient boosted decision trees, particularly XGBoost, are among the most effective methods for tabular data. As deployment in sensitive setti
1/ Excited to report we have a new paper out @Nature today! The bottom line: training data for LLMs does not just fall from the sky - it is created in the context of existing social political institut
AI21 Labs discusses four common agent optimization methods used to balance accuracy against cost and latency constraints. The post indicates the team evaluated all four approaches, likely covering tec
arXiv:2605.11405v1 Announce Type: new Abstract: Data curation has shifted the quality-compute frontier for language-model and contrastive image-text pretraining, but its role for vision-language model
$20/month. Builder: Pinecone's new plan for teams that have outgrown the free tier but aren't running production workloads at scale yet. The prototype-to-production gap has always been an awkward plac
arXiv:2605.11367v1 Announce Type: new Abstract: Recent advances in visual generative models have highlighted the promise of learning generative world models. However, most existing approaches frame wo
arXiv:2605.12437v1 Announce Type: new Abstract: Retrospective novel view synthesis (NVS) of dynamic scenes is fundamental to applications such as sports. Recent dynamic 3D Gaussian Splatting (3DGS) ap
arXiv:2605.11489v1 Announce Type: cross Abstract: 3D Gaussian Splatting (3DGS) enables high-quality real-time 3D rendering but faces challenges in efficiently scaling to ultra-dense scenes and high-re
4/5 Best-of-N: Run one agent config N times in parallel → select the best trajectory. Leverages LLMs’ non-determinism - but hinges on a good eval mechanism (we use an LLM-as-a-Judge). Can increase acc
arXiv:2605.12027v1 Announce Type: new Abstract: Reconstructing dynamic 4D scenes from monocular videos is a fundamental yet challenging task. While recent 3D foundation models provide strong geometric
5/5 Ensemble: Run multiple distinct agent configs in parallel → select the best trajectory. Leverages success of the portfolio vs single variants (see also: @/karpathy’s LLM Council). Can outperform b
7,000 false positives per square millimeter. The culprit was the lab gloves. University of Michigan researchers just upended a core assumption in microplastics science. Latex and nitrile gloves, worn