model releases

arXiv:2604.18874v1 Announce Type: new Abstract: Tool-integrated agents are deployed on the premise that external tools ground their outputs in reality. Yet this very reliance creates a critical attack

arXiv:2604.19193v1 Announce Type: new Abstract: Despite remarkable progress toward general-purpose video models, a critical question remains unanswered: how far are these models from achieving true mu

How far are we from agents that can self-generate world knowledge? The work proposes an outcome-based reward that measures how much an agent's self-generated world knowledge actually improves its task

Coding agents can update prompts, wire in tools, and change application logic across your codebase in a single run. The hard part isn’t getting the agent to make changes, but... The post How to add an

arXiv:2604.19071v1 Announce Type: new Abstract: Evaluating the writing capabilities of large language models (LLMs) remains a significant challenge due to the multidimensional nature of writing skills

arXiv:2604.19406v1 Announce Type: cross Abstract: Common image editing tasks typically adopt powerful generative diffusion models as the leading paradigm for real-world content editing. Meanwhile, alt

Hugging Face Releases ml-intern: An Open-Source AI Agent that Automates the LLM Post-Training Workflow [The 'AI Intern' that actually ships SOTA models ] This isn't just another ML Research Loop wrapp

arXiv:2604.18847v1 Announce Type: new Abstract: As LM agents gain the ability to execute actions on real computer systems, we need ways to not only prevent harmful actions at scale but also effectivel

I wonder if something happened to radicalize NoVa last year I can't believe Virginia dems got away with being so brazen. The 10-1 map looks absolutely insane and its wildly unfair and the public had f

Image models tend to get much more stuck on a particular direction than text models, requiring clearing the context window fairly often. PerfectSquashBench is my new measure of how image models anchor

arXiv:2507.00439v4 Announce Type: replace Abstract: The ability to accurately align LLMs with diverse population groups on subjective questions would have great value. In this work, we show that addin

arXiv:2604.19298v1 Announce Type: cross Abstract: We introduce IndiaFinBench, to our knowledge the first publicly available evaluation benchmark for assessing large language model (LLM) performance on

arXiv:2509.21080v2 Announce Type: replace-cross Abstract: Advancements in Large language models (LLMs) have enabled a variety of downstream applications like story and interview script generation. How

arXiv:2604.19033v1 Announce Type: cross Abstract: In gradient-based learning, a step size chosen in parameter units does not produce a predictable per-step change in function output. This often leads

In the early days of generative AI, building safe and reliable business tools took massive engineering effort and a high tolerance for trial and error. We helped solve that with Vertex AI, our trusted

OpenAI's Privacy Filter is a tool designed to help users protect sensitive information when using OpenAI's services by automatically detecting and redacting or filtering personally identifiable inform

Traditional data catalogs were built as manual inventories for technical users, focusing on table structures rather than the deep context that AI agents need. When agents lack business semantics and d

OpenAI introduced workspace agents in ChatGPT, a feature that enables AI assistants to perform tasks and take actions within connected workspace applications and tools. These agents can help automate

OpenAI introduced workspace agents in ChatGPT, a new feature that enables shared agents capable of managing complex tasks and extended workflows across multiple tools and team members. These agents ar

Anthropic today quietly (as in silently, no announcement anywhere at all) updated their claude.com/pricing page (but not their Choosing a Claude plan page, which shows up first for me on Google) to ad

Iterable Inc., a customer engagement platform, today announced that it launched an artificial intelligence agent designed to help marketers keep customer interactions relevant as campaigns scale. The

I've been mostly prompting the new model directly via the client.images.generate() API, I have no idea if that rewrites my prompts at all or if it passes them straight to the model - details here http

just a friendly reminder that republicans have a trifecta and could pass a federal ban on gerrymandering *tomorrow* if they wanted to. but they don’t want to. they want to rig the system. and then the

Kimi K2.6 is free on Nous Portal for the next 24 hours Made possible by @vercel's AI Gateway & @Kimi_Moonshot Run 'hermes update', then 'hermes model' and select Kimi K2.6 to try out one of the most i

Kimi K2.6, a model developed by Moonshot AI, has achieved the top ranking on OpenRouter's programming leaderboard, indicating superior performance in code generation and programming-related tasks comp

arXiv:2505.09803v3 Announce Type: replace-cross Abstract: In many applications, we wish to fit a parametric statistical model to a small ensemble of spatially distributed random variables ('fields').

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

arXiv:2604.19464v1 Announce Type: cross Abstract: More than half of the global population struggles to meet their civil justice needs due to limited legal resources. While Large Language Models (LLMs)

arXiv:2604.18897v1 Announce Type: new Abstract: We present a systematic empirical study of prompt engineering for formal mathematical reasoning in the context of the SAIR Equational Theories Stage 1 c

As AI models improve, technical practitioners are increasingly turning to agentic AI tools to build with Google Cloud products, from Firebase and the Gemini API, to BigQuery and GKE. But how can you e

arXiv:2601.22737v2 Announce Type: replace Abstract: The robust safety of Vision-Language Large Models (VLLMs) against joint multilingual and multimodal threats remains severely underexplored. Current

LiteParse, our OSS document parser, is really good at parsing complex PDF layouts, text, and tables into a clean spatial grid. The best part is it doesn't use VLMs or any ML models at all. It's entire

arXiv:2604.19536v1 Announce Type: new Abstract: Recent navigation systems achieve strong benchmark results, yet real-world deployment often remains visibly stop-and-go. This bottleneck arises because

This post likely demonstrates running Qwen2 3.6B or 27B model in GGUF format using llama-server with default specifications, showcasing inference capabilities of quantized open-source models. The comm

arXiv:2604.18803v1 Announce Type: cross Abstract: Vision-Language Models (VLMs) are increasingly deployed in settings where reliable visual grounding carries operational consequences, yet their behavi

LM Performance：With only 27B parameters, Qwen3.6-27B outperforms the Qwen3.5-397B-A17B (397B total / 17B active, ~15x larger!) on every major coding benchmark — including SWE-bench Verified (77.2 vs.

arXiv:2604.19405v1 Announce Type: new Abstract: Automatic evaluators such as reward models play a central role in the alignment and evaluation of large vision-language models (LVLMs). Despite their gr

arXiv:2604.19445v1 Announce Type: new Abstract: This paper presents a review for the LoViF Challenge on Real-World All-in-One Image Restoration. The challenge aimed to advance research on real-world a

arXiv:2601.11913v2 Announce Type: replace-cross Abstract: Effectively processing long contexts remains a fundamental yet unsolved challenge for large language models (LLMs). Existing single-LLM-based

OpenAI describes improvements and adaptations made to ChatGPT to better serve clinical and healthcare professionals. The work likely addresses how the model can be made more reliable, accurate, and us

arXiv:2604.18779v1 Announce Type: new Abstract: Existing web agents typically initiate exploration from the root URL, which is inefficient for complex websites with deep hierarchical structures. Witho

Google LLC has emerged as the only cloud “hyperscaler” with a leading frontier artificial intelligence large language model – Gemini – and today it issued a raft of announcements designed to capitaliz

arXiv:2604.18970v1 Announce Type: new Abstract: We can often verify the correctness of neural network outputs using ground truth labels, but we cannot reliably determine whether the output was produce

🚀 Meet Qwen3.6-27B, our latest dense, open-source model, packing flagship-level coding power! Yes, 27B, and Qwen3.6-27B punches way above its weight. 👇 What's new: 🧠 Outstanding agentic coding — surpa

Michele Catasta (@pirroh) at Google Cloud Next talking about what's new with Gemini, agentic applications, and what's next for builders. Fresh off Replit being named Google Cloud Partner of the Year.

arXiv:2604.19642v1 Announce Type: new Abstract: Edge devices such as smartwatches and smart glasses cannot continuously run even the smallest 100M-1B parameter language models due to power and compute