1471
What happens if we remove 50 percent of Llama?
Discover Sparse Llama: A 50% pruned, GPU-optimized Llama 3.1 model with 2:4 sparsity, enabling faster, cost-effective inference without sacrificing accuracy.
Topics /
AI & Machine Learning
The most popular blog posts about ai & machine learning on Hacker News. 3,240 posts.
1472
Next-generation AI models
We've put together an index of testing terms and articles, covering many of the basics of testing and definitions for common searches.
1473
Step 3.5 Flash – Open-source foundation model, supports deep reasoning at speed
Scores represent the mean of the following eight benchmarks listed below, excluding xbench-DeepSearch. The Step 3.5 Flash score is derived under standard settings (i.e., $w/o$ Parallel Thinking).
1474
Trinity large: An open 400B sparse MoE model
A deep dive into the open weight Trinity Large model, covering architecture, sparsity, training at scale, and why we shipped Preview, Base, and TrueBase checkpoints.
1475
Ray Tracing Essentials, Part 1: Basics of Ray Tracing
Ray Tracing Essentials is a seven-part video series hosted by the editor of Ray Tracing Gems, NVIDIA’s Eric Haines.
1476
Wide and Deep Learning: Better Together with TensorFlow
Posted by Heng-Tze Cheng, Senior Software Engineer, Google ResearchThe human brain is a sophisticated learning machine, forming rules by memorizing...
1477
1478
Which table format do LLMs understand best?
We benchmarked 11 data formats, including markdown tables, JSON, CSV, YAML, XML, and more, to see which LLMs understand best. We found surprising accuracy and token cost trade-offs.
1479
The ways we contain Claude across products
As agents grow more capable, so does their potential blast radius. Here's what we learned building containment for claude.ai, Claude Code, and Cowork.
1480
You only need the frontier model for one single edit
/prewalk hands the cheap model a trajectory, not a plan, retaining 92–97% of frontier pass rate at roughly half the cost on SWE-Bench Pro.
1481
AI’s Instagram Problem
AI has an Instagram problem. Just as Instagram’s parade of perfect physiques makes many people feel they don’t measure up, AI’s parade of exciting...
1482
Building a license plate reader from scratch with deep learning
Learn about attention mechanisms and how they are applied for text recognition tasks. We will also use tensorflow attention ocr to train our own number plate reader.
1483
1484
Making a racist AI without really trying (2017)
A cautionary tutorial. Let's make a sentiment classifier!¶Sentiment analysis is a very frequently-implemented task in NLP, and it's no surprise. Recognizing whether people are expressing positi
1485
Our LLM-controlled office robot can't pass butter
Can LLMs control robots? We answer this by testing how good models are at passing the butter – or more generally, do delivery tasks in a household setting. State of the art models struggle, with the best model scoring 40% at Butter-Bench, compared to 95% for humans.
1486
Long context prompting for Claude 2.1
Claude 2.1 excels at retrieving information across its 200K context window, with a simple prompt adjustment improving accuracy from 27% to 98%. Claude 2.1’s performance when retrieving an individual sentence across its full 200K token context window.
1487
SimpleQA
A factuality benchmark called SimpleQA that measures the ability for language models to answer short, fact-seeking questions.
1488
DeepSearcher: A local open-source Deep Research
In contrast to OpenAI’s Deep Research, this example ran locally, using only open-source models and tools like Milvus and LangChain.
1489
1490
Amazon’s Face Recognition Falsely Matched 28 Members of Congress with Mugshots
Amazon’s face surveillance technology is the target of growing opposition nationwide, and today, there are 28 more causes for concern. In a test the ACLU recently conducted of the facial recognition tool, called “Rekognition,” the software incorrectly matched 28 members of Congress, identifying them as other people…
1491
1492
MiniMax M2.1: Built for Real-World Complex Tasks, Multi-Language Programming
在 M2.1 中, 我们系统性提升了 Rust / Java / Golang / C++ / Kotlin / Objective-C / TypeScript / JavaScript 等语言的能力, 多语言任务整体表现达到业内领先水平, 覆盖从底层系统到应用层开发的完整链路。
1493
AI model detects asymptomatic Covid-19 infections through phone-recorded coughs
An artificial intelligence model can detect people who are asymptomatic with Covid-19, through cellphone-recorded coughs. The work was led by Brian Subirana and colleagues at the MIT Auto-ID Lab.
1494
Mxc: Microsoft Execution Containers version 1.0.0
Agents are unlocking enormous productivity gains for customers, but their ability to work across files, networks, and applications can introduce new security risks. This can leave customers feeling like they only have two choices: give agents unrestr
1495
Sharing new research, models, and datasets from Meta FAIR
Meta FAIR is releasing several new research artifacts. Our hope is that the research community can use them to innovate, explore, and discover new ways to apply AI at scale.
1496
1497
Generative AI is overrated, long live old-school AI
TLDR; Don't be dazzled by generative AI's creative charm! Predictive AI, though less flashy, remains crucial for solving real-world challenges a
1498
Claude Platform on AWS
The Claude Platform on AWS is now generally available, offering a new way for AWS customers to access the full set of Claude platform features with AWS authentication, billing, and commitment retirement.
1499
FLUX.2 [Klein]: Towards Interactive Visual Intelligence
Introducing FLUX.2 [klein], our fastest image models. Unified generation and editing in under a second, on hardware you already own. Apache 2.0 on 4B.
1500
Grok-2 Beta Release
We announce our new Grok-2 and Grok-2 mini models.