Impact of Pretraining Word Co-Occurrence on Compositional Generalization In

Comments (1)

badmonster · 12h ago

a subtle but powerful insight: large multimodal models like CLIP don’t just learn individual concepts. they also depend heavily on how often those concepts appear together during training.

Science Fiction, the Future, and Now: Some Mid-Life Reflections (fafnir.journal.fi)

Show HN: XUtil – 40 fast, privacy-friendly developer tools (no ads, no fluff) (xutil.in)

Longevity Might Be All in Your Head (nautil.us)

White-sounding names get called back for jobs more than Black ones, study finds (npr.org)

AI therapy bots fuel delusions and give dangerous advice, Stanford study finds (arstechnica.com)

The Great Exhibition of 1851 (thegazette.co.uk)

Indeed, Glassdoor to lay off 1,300 staff amid AI push (techcrunch.com)

Depopulation won't stop climate change (cremieux.xyz)

curl Cybersecurity Risk Assessment Request (daniel.haxx.se)

PocketBase to OpenAPI Converter (ziadsafwat.github.io)

Ask HN: What do you do with your list of articles links?

Oliver Heaviside: Pioneer of Electromagnetism and Vector Calculus (allaboutcircuits.com)

Even during the war, Ukraine is contributing to the global history of AI (en.wikipedia.org)

Even Realities G1 Smart Glasses Review: Superb Display, but Slow Info (2024) (wired.com)

AI Startups are just Big Techs low cost L&D department (centreforaileadership.org)

What Could a Healthy AI Companion Look Like? (wired.com)

BootLoop: AI Agent for Hardware Software (bootloop.ai)

Code search 40% faster for 100M+ line codebases with quantized vector search (augmentcode.com)

Dismantling Public Values, One Data Center at the Time (helsinki.fi)

France launches criminal investigation into Musk's X over algorithm manipulation (politico.eu)

Sport and longevity: an observational study of international athletes (pubmed.ncbi.nlm.nih.gov)

Intel bombshell: Chipmaker will lay off 2,400 Oregon workers (oregonlive.com)

An Oral History of Marge vs. the Monorail, the Episode That Changed the Simpsons (2020) (vice.com)

Playbook for Building Secure Cloud or Kubernetes Applications (medium.com)

$500K stolen via fake AI coding extension (securelist.com)

The Geological Sublime (harpers.org)

The Salt Path Controversy (theguardian.com)

Study in the UK: The Ultimate Guide for a Master in 2025 (study-from-here.com)

Show HN: Built Podflyy.com, a NoteBookLM Alternative

Psilocybin Shows Promise as Anti-Aging Therapy (neurosciencenews.com)

Show HN: OS Yamato – A gentle digital OS inspired by wabi-sabi (github.com)

Ask HN: If I write a book, is there a legal way I can stop LLMs from reading it?

Scanning for Post-Quantum Cryptographic Support (anvilsecure.com)

Software Freedom Conservancy Files Motion for Summary Adjudication in Vizio Case (sfconservancy.org)

Why the Tech Debt vs. Tech Asset Metaphor Still Holds (michaelbastos.com)

Show HN: DigitalToolPad – privacy-first, locally-run developer tool suite (digitaltoolpad.com)

Show HN: EdgeLLM – 3 offline LLMs for iOS in one SwiftPM line (github.com)

Skills and tech: hydrography's balancing act (hydro-international.com)

Difftastic: A structural diff that understands syntax (github.com)

Calibre 8.6.0 Delivers Dramatic Database Speed Boost (omgubuntu.co.uk)

Threads is nearing X's daily app users, new data shows (techcrunch.com)

Linux Containers – Incus (linuxcontainers.org)

Florida's Brightline Train Defers Bond Payment as Woes Mount (bloomberg.com)

Ask HN: Pocket is dead, what are the best alternative services?

Tesla sales rise in Norway and Spain, boosted by Model Y (reuters.com)

Trace LLM workflows at your app's semantic level, not at the OpenAI API boundary (engineeringideas.substack.com)

Before Tragedy, Texas Repeatedly Rejected Pleas for Flood Alarm Funding (nytimes.com)

King Charles's fiery gathering that shone a light on his beliefs (bbc.com)

What does it take for a software engineer to become an AI Engineer? (mltrenches.substack.com)

Predicting Competitive Pokémon VGC Leads Using Latent Semantic Analysis (jgeekstudies.org)

Impact of Pretraining Word Co-Occurrence on Compositional Generalization In

Comments (1)