Knowledge catalogue
Sam Altman discusses enterprise concerns about AI costs, highlighting that a 5.6 SOL rate represents significant progress in improving cost-efficiency metrics for AI task execution. The post reference
🤗 we just opensourced our tiny beast on @huggingface >0.9B ASR model >Apache license 2.0 >128k long-context transcription >Up to ~90-min audio input >Speaker labels + timestamps in one generation >Mul
Perplexity conducted a cost efficiency comparison of language models, using GLM 5.2 as the baseline metric for cost per task. Results showed GLM 5.2 with an advisor achieved 2.1x cost efficiency on th
We shipped 9k Reachy Minis. They generate 15k hours of conversation a month. On GPT-realtime, that would cost 45k per month. So we built our own: fully open, one line to migrate, 0.25/hour. Free on yo
We stand at a critical crossroads in the debate over AI governance in the United States, and it feels like we are inching closer to a very serious battle over whether or not open source models will ev
arXiv:2607.07127v1 Announce Type: cross Abstract: Lattice field theory is the workhorse of non-perturbative physics, used to simulate phenomena from the strong nuclear force to critical phenomena in m
We're hosting a meetup on agent memory and wikis July 28th at our SF office! Come hear @jacobtpl and myself talk about the frontier research going on in these areas right now. https://luma.com/mylwoab
We're releasing a research preview of a new orchestrator model in Perplexity Computer. The model is an adapted version of GLM 5.2, post-trained for the Computer harness. It delivers near-frontier perf
Sam Altman shared thoughts on the characteristics or qualities that define effective video content. The post likely discusses principles for creating compelling videos, whether from a technical, story
arXiv:2607.06799v1 Announce Type: cross Abstract: Evaluating uncertainty in AI-generated SQL queries requires estimating whether a query is correct, where correct means it executes to the same result
arXiv:2510.13817v2 Announce Type: replace Abstract: The growth of IoT devices in shared environments has outpaced our ability to identify them, posing urgent risks to privacy, safety, and accountabili
arXiv:2607.06807v1 Announce Type: cross Abstract: While enabling effective collaboration on complex tasks, LLM-based Multi-Agent Systems (MAS) face critical security challenges due to vulnerabilities
arXiv:2607.06595v1 Announce Type: cross Abstract: Personal AI agents powered by large language models can reason and act using available tools to access emails, manage calendars, and push code to remo
arXiv:2607.06630v1 Announce Type: new Abstract: Formal robustness certificates for embedded neural-interface models can pass while task accuracy collapses: at perturbation budget e=0.25, EEGNet classi
arXiv:2506.19348v2 Announce Type: replace Abstract: Training-free motion customization imposes motion patterns from reference videos onto video generators through test-time computation. Most existing
arXiv:2607.06634v1 Announce Type: new Abstract: Compact networks built from Clifford algebra Cl(3,0) primitives are exactly SO(3)-equivariant and learn synthetic 3D vector laws from few samples. We as
arXiv:2607.06720v1 Announce Type: new Abstract: Training large language models (LLMs) with extended reasoning has enabled in-context search, in which models iteratively generate, critique, and revise
arXiv:2607.07395v1 Announce Type: cross Abstract: Reliable confidence estimation remains a key limitation of test-time adaptation in vision-language models (VLMs), where prompt tuning improves zero-sh
arXiv:2606.19042v2 Announce Type: replace-cross Abstract: In vibe coding, an emerging AI-driven paradigm, an LLM generates an entire program from a natural language prompt, but what happens to the var
arXiv:2607.06616v1 Announce Type: cross Abstract: Label skew in federated learning (FL) causes client drift and degrades global accuracy. Synthetic data augmentation can reduce this imbalance; however
arXiv:2607.07471v1 Announce Type: cross Abstract: Machine learning models are increasingly deployed in high-stakes domains, raising concerns about both privacy and fairness. Differential Privacy (DP)
While the writing style of LLMs is still as recognizable as ever, a new trend is that humans have started organically writing like them, too (which makes sense: of course you would end up imitating th
'Whoever Jerry is, he was excellent.' That's a customer talking about an agent. @PodiumHQ's Walker Ward sat down with our COO @j_schottenstein to share how LangGraph + LangSmith helped his team take t
arXiv:2607.07216v1 Announce Type: new Abstract: Deepfake (DF) technology poses a significant threat to information integrity, driving the need for robust detection methods. Most DF detectors only cons
arXiv:2607.07123v1 Announce Type: new Abstract: Segmenting slender curvilinear structures such as retinal vessels, cracks, and roads demands topological correctness, as even a single-pixel discontinui
arXiv:2607.06838v1 Announce Type: new Abstract: Humans can navigate an unfamiliar city and gradually form a coherent spatial mental map spanning tens of square kilometers. Can AI build spatial represe
This post documents a Wordle game result (puzzle #1,845) played by Anthropic, showing the progression of guesses through color-coded tile feedback until reaching the correct five-letter word solution
Yann LeCun claimed on social media [7] that my foundational 1990 paper on Neural World Models [1] 'was never accepted through peer review.' This is simply false. The core concepts from the tech report
Yes, AI has definitely made the cheating problem (including getting 'help,' not cheating) worse but the problem was bad already Doing homework improved final test grades for 86% of college students st
Yesterday we launched SWE-1.7 built on the open-source Kimi K2.7. Concerns about Chinese base models are real: K2.7 completed 87% of tasks that other models refuse over human-rights concerns. We train
Vercel announced integration support allowing developers to deploy applications built with Lovable directly to the Vercel platform. This integration streamlines the deployment workflow by connecting L
You've heard of Infrastructure as Code- but agent evals can now ride your existing Terraform setup! I've been using the new LangSmith Terraform provider to auto-provision online evals + monitoring ale
arXiv:2510.24108v2 Announce Type: replace-cross Abstract: Human demonstrations are widely considered the cornerstone of end-to-end (E2E) autonomous driving despite human demonstration's scarcity for l
arXiv:2510.17476v2 Announce Type: replace Abstract: Equitable access to reliable health information is vital when integrating AI into healthcare. Yet, information quality varies across languages, rais
1/ GPU capacity is getting more distributed. Data often isn’t. Teams can find GPUs across clouds, Kubernetes, Slurm, or on-prem clusters, but then still have to move models, datasets, and checkpoints
Physical AI and robotics are moving beyond impressive demonstrations into a new phase of practical deployment, with companies now targeting specific, high-value use cases in manufacturing and logistic
1/3 Best-of-N leaves $$ on the table by not accounting for variance in task difficulty. We built budget-aware execution: turn the dial on compute or speed, while keeping quality constant, to save cost
2/3 By building a reliable early stopping mechanism, we could apply cascading (save up to 44% compute by not running unnecessary rollouts ) or parallel execution (up to 25% faster by sparing wait time
3/3 Best-of-N is too flat for real world variance in task difficulty. Budget-aware execution turns compute & speed into dials you tune per workload. Full write-up here: https://www.ai21.com/blog/impro
arXiv:2607.06115v1 Announce Type: cross Abstract: In next-generation networks, communication systems will no longer be limited to data transmission and will be expected to acquire awareness of the sur
arXiv:2607.05615v1 Announce Type: new Abstract: Activation steering via sparse autoencoders (SAEs) enables behavioral control of large language models without task-specific fine-tuning, but standard m
arXiv:2512.07997v2 Announce Type: replace-cross Abstract: Gestures are an integral part of our daily interactions with the environment. Hand gesture recognition (HGR) is the process of interpreting hu
arXiv:2601.17257v2 Announce Type: replace Abstract: We introduce a constrained optimization framework for training transformers that behave like optimization descent algorithms. Specifically, we enfor
arXiv:2607.06252v1 Announce Type: cross Abstract: Many problems in science and engineering are difficult to model accurately, either due to unknown physical mechanisms, poorly quantified measurement u
arXiv:2607.06401v1 Announce Type: new Abstract: World models -- internal simulators that learn the structure and dynamics of an environment -- have become one of the most actively debated concepts in
arXiv:2511.17240v3 Announce Type: replace-cross Abstract: We study the problem of learning an unknown graph via group queries on node subsets, where each query reports whether at least one edge is pre
Replit is offering a free custom domain for applications built on their platform, with no additional cost through July 17. Users can build and publish their apps, then assign a custom domain name as p
arXiv:2607.06382v1 Announce Type: cross Abstract: A persistent empirical observation is that trained neural networks outperform their neural tangent kernel (NTK) limit on tasks with compositional stru
arXiv:2210.16286v2 Announce Type: replace Abstract: To understand the training dynamics of neural networks, prior studies have considered the mean-field limit of two-layer neural networks as the width
arXiv:2602.11059v2 Announce Type: replace-cross Abstract: This paper addresses the issue of inversion in cases where (1) the observation system is modeled by a linear transformation and additive error
GPT-Live is OpenAI's voice model designed to be an engaging conversational partner with natural speech and a distinct personality. The model prioritizes making interactions enjoyable for users through
arXiv:2607.05406v1 Announce Type: cross Abstract: Large language models generate code from natural language prompts, enabling 'vibe coding,' which allows non-programmers to develop computational solut
CNBC: A look at Chinese lidar maker Hesai, blacklisted by the US DOD in 2024, as it expands in the US; Hesai says it has ~33% of the global automotive lidar market — Robots on the factory floor. Self-
arXiv:2602.11391v4 Announce Type: replace Abstract: Objective: This study develops and validates a patient simulation framework that aligns with the National Institute of Standards and Technology (NIS
arXiv:2607.06479v1 Announce Type: new Abstract: Physics-informed neural networks (PINNs) provide a promising framework for solving partial differential equations while embedding the underlying physica
arXiv:2607.05467v1 Announce Type: new Abstract: Fog severely degrades the visibility of small unmanned aerial vehicles (UAVs) in skydominant, long-range imagery, reducing the reliability of downstream
arXiv:2606.13566v2 Announce Type: replace Abstract: Current discussions of AI in scientific discovery are often dominated by two visible capabilities: search over existing knowledge and execution thro
arXiv:2607.06214v1 Announce Type: new Abstract: This paper offers a toy framework for considering curiosity as an ecosystem. First, it suggests that a single agent's inquiry policy (how, when, and why
arXiv:2607.06478v1 Announce Type: new Abstract: Traffic signs are crucial components of road safety, serving as visual tools under all lighting conditions. The Manual on Uniform Traffic Control Device
arXiv:2607.05434v1 Announce Type: cross Abstract: Artificial Intelligence (AI) models, at their core, apply general learnings from broad datasets to individual circumstances using probabilistic behavi