research

arXiv:2605.07271v1 Announce Type: cross Abstract: Layer pruning efficiently reduces Large Language Model (LLM) computational costs but often triggers sudden performance collapse. Existing representati

arXiv:2605.07359v1 Announce Type: new Abstract: Compared to RGB images, raw sensor data provides a richer representation of information, which is crucial for accurate recognition, particularly under c

arXiv:2603.09742v2 Announce Type: replace Abstract: Neural oscillators that originate from second-order ordinary differential equations (ODEs) have shown competitive performance in learning mappings b

arXiv:2605.08070v1 Announce Type: new Abstract: A standard technique for scaling inference-time reasoning is Self-Consistency, whereby multiple candidate answers are sampled from an LLM and the most c

arXiv:2605.07385v1 Announce Type: cross Abstract: Editing a 3D asset locally, modifying a target region while preserving the rest, is a fundamental requirement of native 3D editing. Existing methods e

arXiv:2605.07552v1 Announce Type: new Abstract: The rapid advances in deep learning have significantly enhanced the accuracy of multimodal 3D human pose estimation (HPE). However, the state-of-the-art

arXiv:2605.06765v1 Announce Type: cross Abstract: Human speech conveys expressiveness beyond linguistic content, including personality, mood, or performance elements, such as a comforting tone or humm

arXiv:2605.07451v1 Announce Type: new Abstract: Neural network verification is an active and rapidly maturing research area, with a growing ecosystem of solvers and tools. The VNN-LIB standard was int

arXiv:2605.07367v1 Announce Type: cross Abstract: Cameras and LiDAR degrade in rain, fog, and snow, while millimeter-wave radar remains largely unaffected. We align a radar encoder to frozen SigLIP vi

arXiv:2605.07522v1 Announce Type: new Abstract: Accurate weather forecast reporting enables individuals and communities to better plan daily activities and agricultural operations. However, the curren

arXiv:2605.07915v1 Announce Type: new Abstract: Tokenizers are a crucial component of latent diffusion models, as they define the latent space in which diffusion models operate. However, existing toke

What will we be like when he is gone? Can we return to mutual respect? Can we believe we are all on the same team as Obama and McCain did? Can we imagine the mutual respect of those two, competitors b

arXiv:2605.07969v1 Announce Type: new Abstract: Diffusion models perform remarkably well on high-dimensional data such as images, often using only a modest number of reverse-time steps. Despite this p

arXiv:2605.06723v1 Announce Type: new Abstract: Language models often generate reasoning before giving a final answer, but the visible answer does not reveal when the model's answer preference became

arXiv:2602.09229v3 Announce Type: replace Abstract: Cosine similarity normalizes both sides; dot product normalizes neither. We propose a 2x2 framework that independently controls query-side and docum

arXiv:2605.07756v1 Announce Type: cross Abstract: Modern deep models are often pretrained on large-scale data with missing labels using composite objectives, where the relative weights of multiple los

arXiv:2605.07533v1 Announce Type: new Abstract: Large Language Models (LLMs) have recently demonstrated strong performance in machine translation (MT). However, most prior work focuses on improving or

arXiv:2602.01752v2 Announce Type: replace Abstract: As large language models (LLMs) generate increasingly human-like text, watermarking has emerged as a promising solution for reliable attribution bey

arXiv:2605.06913v1 Announce Type: cross Abstract: We present You Only Stack Once (YOSO), an automated pipeline designed to detect faint, slow-moving Solar System objects in wide-field astronomical sur

arXiv:2605.05405v2 Announce Type: replace Abstract: Semantic search of Earth observation archives remains challenging. Visual foundation models such as CLAY produce rich embeddings of satellite imager

A Washington Post/Ipsos poll found that 71% of respondents said Donald Trump is not honest or trustworthy, and 67% said he does not carefully consider important decisions. The poll data was shared by

Back from a little family break! Lots has happened, and I’m planning to do a deeper dive into the most interesting architectural components (soon). Btw, are there any major architectures I missed belo

France is the only European country that turned nuclear generation into a structural competitive advantage. 57 reactors built between the 1970s and 1990s produce 70% of its electricity today. Wholesal

This statement by François Chollet argues that true understanding of an idea requires the ability to formalize it mathematically or through code, rather than merely having intuitive grasp. The premise

(INTENTIONALLY) LOST IN TRANSLATION: Democrats: We'd like cops to stop killing minorities. Republicans: Dems hate police. Democrats: Women should have the right to choose. Republicans: Dems want to ki

François Chollet argues that intuition serves as the foundational precursor to the development of significant ideas and innovations. This perspective suggests that before formal analysis, rigorous met

防衛分野で働くSoftware Engineerのインタビュー記事を公開しました🐟 https://sakana.ai/defense-swe-interview-2026/ 内容を一部ご紹介 ・指揮統制システムや偽情報対策技術の開発など、防衛分野のSoftware Engineerが担う役割 ・多様な技術領域をまたいだ開発スタイル ・ミッションクリティカルな領域での緊張感と充実感 防衛×AIの最

The Top AI Papers of the Week (May 4 - 10) - Conductor - HeavySkill - Horizon Generalization - 1,000 Synthetic Computers - Self-Improving Pretraining - Coordination as Architecture - Connect Four Alph

Nous Research announced that a promotional code or giveaway has been claimed by new users and is no longer available. The organization encouraged users who missed the offer to enable notifications on

Europe does not lack innovation. It lacks scale. European universities produce world-class research, engineers and technology. But too many companies remain trapped inside fragmented national markets

Francois Chollet argues that AI amplifies existing disparities in user agency, where individuals with high agency gain further capabilities while those with low agency become increasingly dependent, c

Mark: 1/ First milestone: the Physical Turing Test. You literally can’t tell if a human or robot is doing the task. 2/ Next: Physical API. A fleet of robots, configured like software via APIs & CLI. 3

Need inexpensive API? I'm already a user of the Nous portal, but if you're not, you might want to take advantage of this deal quickly now before the slots fill up. Ready to finally make the switch? He

Nous Research announced a congratulatory message regarding a collaboration or achievement involving OpenRouter, likely related to AI model deployment or integration. The post celebrates a milestone or

One theorem every ML engineer should know: The Johnson–Lindenstrauss Lemma. It states that high-dimensional data can be projected into a much lower-dimensional space while approximately preserving pai

Reproducing all of Schmidhuber’s papers (1990-2025) using an AI coding assistant. Cool project by @yaroslavvb! It even reproduced the “World Models” paper by me and @SchmidhuberAI with a toy env, with

At Apple, we believe privacy is a fundamental human right. As AI capabilities increase and become more integrated into people’s daily lives, advancing research in privacy-preserving techniques is incr

Microsoft Research is excited to release an open dataset of approximate transmission topology of the U.S. power grid derived from publicly available data. The ability to study transmission-level power

Mechanistic interpretability research aims to uncover specific neurons and circuits in neural networks responsible for tasks, but over a decade of efforts suggests these findings may not translate int

For those interested, I will be doing a live session on this topic soon: https://academy.dair.ai/events/cmovobp97000904l5h0n9a2yz Sign up if you are interested in some of the tools we are releasing so

Technology is changing the way we make babies. The pioneering work of the scientists who invented IVF led to the birth of the first “test tube baby” in 1978. We’ve come a long, long way since then. Th

MIT Technology Review Explains: Let our writers untangle the complex, messy world of technology to help you understand what’s coming next. You can read more from the series here. Eight passengers aboa

Jim is always a crowd favorite at AI Ascent. His ability to simplify the latest research into a clear 'what and why it matters' while adding humor along the way is unmatched. If you're interested in p

LLM Wikis + HTML Artifacts are insanely powerful. You should seriously consider this in your workflows. LLM Wikis captures all the important information that lets you and your agents do meaningful wor

In the second week of the landmark trial between Elon Musk and OpenAI, Musk’s motivations for bringing the suit were under scrutiny. Last week, Musk took the stand, alleging that OpenAI CEO Sam Altman

Nous Research announced upcoming native Windows support, likely indicating plans to optimize their AI models or tools for Windows operating systems natively rather than through workarounds or alternat