Login

Recent searches

No recent searches

Search options

Only available when logged in.
masto.ai is one of the many independent Mastodon servers you can use to participate in the fediverse.
A general Mastodon server for all languages.

Administered by:

Server stats:

2.3K
active users

masto.ai: AboutStatusProfiles directoryPrivacy policy

Mastodon: AboutGet the appKeyboard shortcutsView source codev4.3.3

#research

261 posts223 participants5 posts today
Public
ndocist

Preprint: “The Economic Impact of Open Science: A Scoping Review”
"Evidence, although limited, indicates that #openscience accelerates #research processes, reduces the related costs, fosters #innovation by improving access to data and resources and this ultimately generates #economicgrowth "
infodocket.com/2025/03/01/prep

Library Journal infoDOCKETPreprint: “The Economic Impact of Open Science: A Scoping Review”The preprint (ver.1) linked below was recently shared  on MetaArXiv. Title The Economic Impact of Open Science: A Scoping Review Authors Lena Tsipouri National and Kapodistrian University of Athens, Greece OPIX PC, Greece Sofia Liarti OPIX PC, Greece Silvia Vignetti CSIL, Italy Izabella Martins Grapengiesser Technopolis Consulting Group Source MetaArXiv February 21, 2025 Abstract This […]
Public *
rexi

medicalxpress.com/news/2025-02

…people have different brain structures that never quite match when aligned to a standardized #brainatlas…different input dimensions are required for each subject.

…neuroscience-informed activity mapping within the #fMRI encoder…allows the system to accommodate these varying input shapes across subjects.

By separating a voxel's functional information from its raw fMRI value, the model leverages pre-existing knowledge from #neuroscience #research…

Medical Xpress · Direct translation of brain imaging to text with MindLLMBy Justin Jackson
#MindLLM
Public
Kim Perales

“Medical research in the US:

Doesn't look so good -& it gets worse.

Before grants are awarded -evaluated in meetings called study sections.

Before a study section can meet -listed in the fed registry for at least 15 days.

Meetings can't even be scheduled -none of the committed funds are going out.

~30OK scientists🚨are wondering if they are going to keep their jobs, & the most vulnerable are the students & apostdocs.”
-M Johnson

#Research: DECIMATED

Cumulative award count from the NIH for the year:
Graph depicting cumulative NIH awards year-to-date (YTD) by award notice date from Jan to Feb. The x-axis shows the date, while the y-axis displays the cumulative number of awards. Multiple lines represent data for different years…
#USPol#DOGE#NIH
Public
N-gated Hacker News

An academic snoozefest where scientists brag about making #AI smarter by "improving" how it pays attention—because clearly, that's the only thing holding it back. 🤖📚 Meanwhile, we're still waiting for the day when AI can pay attention to our emails and reply with anything more than a 🤔.
arxiv.org/abs/2502.12962 #Research #AcademicConference #AttentionMechanisms #AIHumor #TechCritique #HackerNews #ngated

arXiv logo
arXiv.orgInfinite Retrieval: Attention Enhanced LLMs in Long-Context ProcessingLimited by the context window size of Large Language Models(LLMs), handling various tasks with input tokens exceeding the upper limit has been challenging, whether it is a simple direct retrieval task or a complex multi-hop reasoning task. Although various methods have been proposed to enhance the long-context processing capabilities of LLMs, they either incur substantial post-training costs, or require additional tool modules(e.g.,RAG), or have not shown significant improvement in realistic tasks. Our work observes the correlation between the attention distribution and generated answers across each layer, and establishes the attention allocation aligns with retrieval-augmented capabilities through experiments. Drawing on the above insights, we propose a novel method InfiniRetri that leverages the LLMs's own attention information to enable accurate retrieval across inputs of infinitely length. Our evaluations indicate that InfiniRetri achieves 100% accuracy in the Needle-In-a-Haystack(NIH) test over 1M tokens using a 0.5B parameter model, surpassing other method or larger models and setting a new state-of-the-art(SOTA). Moreover, our method achieves significant performance improvements on real-world benchmarks, with a maximum 288% improvement. In addition, InfiniRetri can be applied to any Transformer-based LLMs without additional training and substantially reduces inference latency and compute overhead in long texts. In summary, our comprehensive studies show InfiniRetri's potential for practical applications and creates a paradigm for retrievaling information using LLMs own capabilities under infinite-length tokens. Code will be released in link.
Public
N-gated Hacker News

Ah yes, the groundbreaking discovery of #narwhals doing... narwhal things, now brought to you by a #drone. 🤖✨ But wait, what's that? A 400 Bad Request? Clearly, the internet gods have deemed this highly 'critical' #research too dangerous to share. 🚫🐋🔧
phys.org/news/2025-02-drone-ca #interneterror #wildlife #conservation #techinnovation #HackerNews #ngated

Phys.org · Drone captures narwhals using their tusks to explore, forage and playBy Florida Atlantic University
Public
Bích-Mây Nguyễn :verified:

"As of this week, some of the agency’s 27 institutes and centers are still issuing no new grants at all, one NIH official told me. Grant-management officers, who sign their name to awards, are too afraid, the official said, that violating the president’s wishes will mean losing their livelihood."

theatlantic.com/health/archive

The Atlantic · Inside the Collapse at the NIHBy Katherine J. Wu
#healthcare#medicine#science
Public
Nick Walker

Overleaf is trying to find out if there's SaaS revenue to be had for using AI to patch over the bad experience of LaTeX.

If you sell bikes with square wheels, do you come out ahead when you advertise your wheel-rounding services?

Overleaf Error Assist Annual, get unlimited AI help to fix LaTeX errors faster: $59
#AI#research
Public
Tim (Wadhwa-)Brown :donor:

Interesting links of the week:

Strategy:

* blog.cryptographyengineering.c - @matthew_d_green with the hard questions on UK approach to E2E encryption

Standards:

* itu.int/rec/T-REC-X.1060-20210 - cyber defense centers, a standards based approach

Threats:

* lrqa.com/en/cyber-labs/the-$1. - cracking crypto heist Satoshi
* linkedin.com/posts/daniel-cuth - backdooring generative systems from @dcuthbert

Exploitation:

* positronsecurity.com/blog/2025 - persisting with k8s

Nerd:

* benjamintoll.com/2022/02/04/on - have you tried nspawn?

#security, #research

A Few Thoughts on Cryptographic Engineering · Three questions about Apple, encryption, and the U.K.Two weeks ago, the Washington Post reported that the U.K. government had issued a secret order to Apple demanding that the company include a “backdoor” into the company’s end-to-e…
ExploreLive feeds
About