Knowledge catalogue
arXiv:2605.08841v1 Announce Type: new Abstract: Vision-Language Models (VLMs) exhibit systematic bias toward visual illusions, recalling memorized facts rather than perceiving actual visual difference
arXiv:2605.08498v1 Announce Type: cross Abstract: We introduce MathConstraint, a hard, adaptive benchmark for evaluating the combinatorial reasoning capabilities of LLMs. We combine constraint satisfa
arXiv:2605.09504v1 Announce Type: cross Abstract: We present swarm-attack, an open-source adversarial testing framework in which multiple lightweight LLM agents coordinate through shared memory, paral
arXiv:2605.09900v1 Announce Type: new Abstract: A vision-language model can look at a knot diagram and report what it sees, yet fail to act on that structure. KnotBench pairs an 858,318-image corpus f
arXiv:2605.10912v1 Announce Type: new Abstract: Large language and vision-language models increasingly power agents that act on a user's behalf through command-line interface (CLI) harnesses. However,
arXiv:2605.06707v1 Announce Type: cross Abstract: This paper presents an eight-week observational comparison of 68 single-file HTML generations collected across 17 public experiments in the 'HTML AI B
Anthropic just shipped sleep into agents. When you sleep, your hippocampus replays the day's neural sequences to the cortex during 150-220 Hz bursts called sharp-wave ripples. The replay runs about 20
Anthropic: Anthropic researchers detail “natural language autoencoders”, which convert LLM activations, the numbers encoding a model's thoughts, into natural language text — When you talk to an AI mod
arXiv:2605.05175v1 Announce Type: cross Abstract: Background: Existing MRI LLM benchmarks rely mainly on review-book multiple-choice questions, where top proprietary models already score highly, limit
arXiv:2605.03034v1 Announce Type: new Abstract: Agentic systems involved in high-stake decision-making under adversarial pressure need formal guarantees not offered by existing approaches. Motivated b
arXiv:2605.01338v1 Announce Type: new Abstract: System-level diagrams encode the architectural blueprint of chip design, specifying module functions, dataflows, and interface protocols. However, non-s
Elon has been the longest and by far the biggest voice actively warning about the dangers of AI for a very long time The entire reason he started OpenAI was this one thing.....to make sure AI is built
arXiv:2605.03217v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly deployed in settings that require nuanced ethical reasoning, yet existing bias evaluations treat model out
omg @bcherny with banger quotes “the future is more async agents… this is why we emphasize verification” “if you’re familiar with higher order functions, routines are higher order prompts” “default is
arXiv:2605.02173v1 Announce Type: new Abstract: We evaluate the long-context retrieval and reasoning capabilities of five frontier large language models with advertised 1M-token context windows on a c
arXiv:2603.18196v3 Announce Type: replace-cross Abstract: Investigating cybersecurity incidents requires collecting and analyzing evidence from multiple log sources, including intrusion detection aler
Two frontier labs. One accelerated computing platform. Congrats to @SpaceX and @AnthropicAI on the new compute partnership, powered by 220,000+ NVIDIA GPUs inside Colossus 1. The future of AI runs on
I recently talked with Joseph Ruscio about AI coding tools for Heavybit's High Leverage podcast: Ep. #9, The AI Coding Paradigm Shift with Simon Willison. Here are some of my highlights, including my
Anthropic announced a major compute partnership with SpaceX involving over 300 MW of new computational capacity and 220,000 NVIDIA GPUs to be brought online within one month, dedicated to powering Cla
We're winding back our peak hours limit reduction and doubling 5 hour limits. Excited to partner with SpaceX to bring you more compute and we'll keep pushing to bring you the best coding agent in the
We’ve agreed to a partnership with @SpaceX that will substantially increase our compute capacity. This, along with our other recent compute deals, means that we’ve been able to increase our usage limi
arXiv:2605.02647v1 Announce Type: new Abstract: Large language models (LLMs) remain vulnerable to jailbreak attacks that bypass safety alignment and elicit harmful responses. A growing body of work sh
arXiv:2605.02363v1 Announce Type: new Abstract: Deployed language models must produce outputs that are both correct and format-compliant. We study this structured-output reliability gap using two math
/elsewhere/sightings/ I have a new camera (a Canon R6 Mark II) so I'm taking a lot more photos of birds. I share my best wildlife photos on iNaturalist, and based on yesterday's successful prototype I
I have been testing DeepSeek-V4-Pro with the Pi coding agent. I am mindblown by how well it works out of the box. A few notes: I spent a few hours building an LLM wiki with an agent powered entirely b
Tool: iNaturalist Sightings I wanted to see my iNaturalist observations - across two separate accounts - grouped by when they occurred. I'm camping this weekend so I built this entirely on my phone us
arXiv:2602.15983v2 Announce Type: replace-cross Abstract: Large language models (LLMs) can translate natural language into optimization code, but silent failures pose a critical risk: code that execut
arXiv:2604.25102v1 Announce Type: new Abstract: Typographic prompt injection exploits vision language models' (VLMs) ability to read text rendered in images, posing a growing threat as VLMs power auto
We’re excited to introduce KAME: Tandem Architecture for Enhancing Knowledge in Real-Time Speech-to-Speech Conversational AI, accepted at #ICASSP2026! 🐢 Blog https://pub.sakana.ai/kame/ Paper https://
arXiv:2512.21110v3 Announce Type: replace Abstract: Current Large Language Models (LLMs) safety approaches focus on explicitly harmful content while overlooking a critical vulnerability: the inability
arXiv:2604.23341v1 Announce Type: cross Abstract: The deployment of Large Language Models (LLMs) as assistants in electric grid operations promises to streamline compliance and decision-making but exp
arXiv:2604.24618v1 Announce Type: new Abstract: We evaluate the propensity of frontier models to sabotage or refuse to assist with safety research when deployed as AI research agents within a frontier
arXiv:2604.23366v1 Announce Type: new Abstract: Autonomous multi-agent LLM systems are increasingly deployed to investigate operational incidents and produce structured diagnostic reports. Their trust
Introducing talkie: a 13B vintage language model from 1930 New project from Nick Levine, David Duvenaud, and Alec Radford (of GPT, GPT-2, Whisper fame). talkie-1930-13b-base (53.1 GB) is a '13B langua
arXiv:2510.12826v2 Announce Type: replace-cross Abstract: As large language model (LLM) agents are deployed autonomously in diverse contexts, evaluating their capacity for strategic deception becomes
Processing loan applications traditionally takes a staggering amount of time going through paperwork - ~dozens of hours every month looking at the loan application and cross-checking it with user-subm
Spot on 🎯 We don't need frontier intelligence to automate searches and sending emails - We don't need trillion parameter models to be able to summarize articles or technical documents - We don't need
Vercel announced the availability of GPT-5.5 through its AI Gateway service, expanding the model options developers can access through the platform. This integration allows developers to leverage Open
Available on @ollama ! 🤝🤝 Qwen 3.6 27B model is available on Ollama! Use it with all the integrations in Ollama or chat with the model. Chat with the model: ollama run qwen3.6:27b OpenClaw: ollama lau
Customer engagement platform company Braze Inc. today announced a new pair of agentic artificial intelligence tools for marketers and launched a new Creative Studio that links design software directly
Microsoft is rolling out a new Agent Mode inside Office apps like Word, Excel, and PowerPoint this week. Previously described by Microsoft as 'vibe working,' the Agent Mode is a more powerful version
arXiv:2604.19811v1 Announce Type: cross Abstract: AI leaders and safety reports increasingly warn that advances in model reasoning may enable biological misuse, including by low-expertise users, while
🚨 OpenAI just launched GPT-5.5. The OpenAI team was nice enough to give me early access over the last several weeks, and I just want to flag: there is a certain class of models (one that we’re hitting
arXiv:2604.19784v1 Announce Type: cross Abstract: Recently, it has been found that frontier AI models can resist their own shutdown, a behavior known as self-preservation. We extend this concept to th
Waiting at Superchargers is rare, but when it happens, customers should be able to plan with confidence. Superchargers are the only fast chargers with predictive wait times and we're now using vehicle
a bunch here where I’m saying ok Garry’s kinda right?! 👀…in some ways :) we’re making this loop much easier to close out of the box soon If more people get into evals & traces to ground self-improving
Could a sequence of poetic gibberish manipulate the world’s most advanced AI? Recent tests on GPT-5 by @ChristophHeilig show it consistently rates nonsensical 'word salad' as higher quality than clear
arXiv:2604.19598v1 Announce Type: cross Abstract: This study compared repeated generation consistency of exercise prescription outputs across three large language models (LLMs), specifically GPT-4.1,
arXiv:2604.19533v1 Announce Type: cross Abstract: We introduce the Cyber Defense Benchmark, a benchmark for measuring how well large language model (LLM) agents perform the core SOC analyst task of th
arXiv:2602.18571v2 Announce Type: replace-cross Abstract: While significant progress has been made in automating various aspects of software development through coding agents, there is still significa
arXiv:2604.18878v1 Announce Type: new Abstract: We introduce LegalBench-BR, the first public benchmark for evaluating language models on Brazilian legal text classification. The dataset comprises 3,10
Having a 20-agent system is a million times more powerful than 100 agents working in silos. Last night, I stayed up way too late fixing and improving my digital workforce for my AI Agent Mastermind pr
I hope this helps. If you need more help let me know: Connecting OpenClaw to the X API is straightforward now thanks to X’s official native support... The best and most direct method uses the official
Scaling up Kimi K2.6 with more NVIDIA Blackwell GPUs! Will be adding more in the coming days. Try it with OpenClaw: ollama launch openclaw --model kimi-k2.6:cloud Try it with Hermes Agent: ollama laun
AI reviewers then ranked the submissions, and gave the same ordering every time, regardless of model doing the ranking: Codex GPT-5.4 > GPT-5.3-Codex > Opus 4.6 > humans. Paper: http://claude-code-eco
arXiv:2604.16270v1 Announce Type: cross Abstract: The complexity of Vietnam's legal texts presents a significant barrier to public access to justice. While Large Language Models offer a promising solu
Vercel's AI Gateway now supports Kimi K2.6, a language model that developers can access through Vercel's unified AI platform. This integration allows users to leverage Kimi K2.6's capabilities alongsi
Kimi K2.6 raises the bar for open-source models. 🦙 available on Ollama's cloud! Try it with OpenClaw: ollama launch openclaw --model kimi-k2.6:cloud Try it with Hermes Agent: ollama launch hermes --mo
Ollama we love you 👨💻🚀 Kimi K2.6 raises the bar for open-source models. 🦙 available on Ollama's cloud! Try it with OpenClaw: ollama launch openclaw --model kimi-k2.6:cloud Try it with Hermes Agent: o
Philosophy (among other things) grad here. I could write a whole essay about this video, and mostly the reactions to it. People are dunking on her because of what she symbolises more than what she say