HackerNews Digest Daily

Subscribe
Archives
June 28, 2023

Hacker News Top Stories with Summaries (June 29, 2023)

    <style>
        p {
            font-size: 16px;
            line-height: 1.6;
            margin: 0;
            padding: 10px;
        }
        h1 {
            font-size: 24px;
            font-weight: bold;
            margin-top: 10px;
            margin-bottom: 20px;
        }
        h2 {
            font-size: 18px;
            font-weight: bold;
            margin-top: 10px;
            margin-bottom: 5px;
        }
        ul {
            padding-left: 20px;
        }
        li {
            margin-bottom: 10px;
        }
        .summary {
            margin-left: 20px;
            margin-bottom: 20px;
        }
    </style>
        <h1> Hacker News Top Stories</h1>
        <p>Here are the top stories from Hacker News with summaries for June 29, 2023 :</p>

    <div style="margin-bottom: 20px;">
        <table cellpadding="0" cellspacing="0" border="0">
            <tr>
                <td style="padding-right: 10px;">
                <div style="width: 200px; height: 100px; border-radius: 10px; overflow: hidden; background-image: url('https://hackernewstoemail.s3.us-east-2.amazonaws.com/hnd2'); background-size: cover; background-position: center;">

XGen-7B, a new 7B foundational model trained on up to 8K length for 1.5T tokens

https://blog.salesforceairesearch.com/xgen/

Summary: Researchers have developed XGen-7B, a series of 7B long language models (LLMs) trained on up to 8K input sequence length for up to 1.5T tokens. XGen-7B achieves comparable or better results on standard NLP benchmarks compared to state-of-the-art open-source LLMs. The model also performs well in text and code tasks, such as MMLU and HumanEval. XGen-7B is designed to handle long sequences, making it suitable for applications like summarizing text, writing code, and predicting protein sequences.

    <div style="margin-bottom: 20px;">
        <table cellpadding="0" cellspacing="0" border="0">
            <tr>
                <td style="padding-right: 10px;">
                <div style="width: 200px; height: 100px; border-radius: 10px; overflow: hidden; background-image: url('https://hackernewstoemail.s3.us-east-2.amazonaws.com/hnd2'); background-size: cover; background-position: center;">

OpenOrca: open source dataset and instruct-tuned LLMs

https://erichartford.com/openorca

Summary: Eric Hartford announces OpenOrca, an open-source dataset and series of instruct-tuned language models inspired by Microsoft's GPT-4 Orca. With the help of an AI/ML engineering team, they created a dataset consisting of ~1 million FLANv2 augmented with GPT-4 completions and ~3.5 million FLANv2 augmented with GPT-3.5 completions. OpenOrca-LLaMA-13b is expected to release in mid-July 2023, with evaluation findings and dataset publication. The team is seeking GPU compute sponsors for training OpenOrca on various platforms.

Want to read the full issue?
Powered by Buttondown, the easiest way to start and grow your newsletter.