29 stories tagged gpt-4, newest first.
331 points by joshwa | 113 comments
arXiv paper introducing SWE-Bench+, an empirical analysis revealing data quality issues in the SWE-bench coding benchmark for LLMs, such as solution leakage and weak tests.
871 points by simonw | 530 comments
Simon Willison's blog post reviewing the major developments, themes and pivotal moments in large language models during 2024.
376 points by majc2 | 86 comments
Blog post evaluating finetuned Mistral, Llama3 and Solar LLMs against OpenAI's GPT-4 on accuracy for structured data extraction from press releases.
382 points by davidbarker | 127 comments
OpenAI research post describing scalable methods to decompose GPT-4's internal representations into 16 million interpretable features.
521 points by mellosouls | 199 comments
SSRN page for a Chicago Booth working paper examining whether large language models can perform financial statement analysis like professional analysts.
350 points by RafelMri | 365 comments
Gary Marcus Substack post from Feb 2024 describing user reports of ChatGPT producing bizarre, nonsensical output and framing it as a warning.
351 points by gregsadetsky | 85 comments
A page with a video demo and code repo link for a real-time GPT-4 vision and voice remake of Google's staged Gemini demo.
324 points by dmakian | 89 comments
Substack post by David Hershey sharing tips, examples, and lessons from fine-tuning Mistral 7B to make Magic: The Gathering draft picks.
455 points by simonw | 196 comments
Simon Willison's blog post exploring OpenAI's custom GPTs feature, how to configure them, and examples of GPTs he built.
572 points by hackerbeat | 281 comments
An X post by Javi Lopez describing how he built an Angry Birds clone, "Angry Pumpkins," using GPT-4 for code and Midjourney/DALL-E for graphics.
793 points by rushingcreek | 316 comments
Phind announces its own coding-focused LLM, built on CodeLlama-34B fine-tunes, claiming GPT-4-level coding answers at 5x speed with 16k context.
327 points by shishirpatil | 78 comments
GitHub repository for MemGPT, an open-source project that gives LLMs self-editing memory management to handle unbounded context.
738 points by spdustin | 366 comments
A GitHub markdown file in the ChatGPT-AutoExpert repo documenting the hidden system prompts ChatGPT uses with GPT-4, including DALL·E and other modes.
825 points by kcorbitt | 158 comments
Show HN post sharing notebooks and lessons on fine-tuning Llama 2 7B to match GPT-4 on a recipe classification task at far lower cost.
737 points by rushingcreek | 258 comments
Phind blog post announcing fine-tuned CodeLlama-34B models that score 67.6% and 69.5% pass@1 on HumanEval, beating GPT-4's reported 67%.
360 points by 152334H | 162 comments
Blog post arguing that GPT-4's non-deterministic outputs at temperature zero are caused by its sparse mixture-of-experts architecture and batched routing.
569 points by bx376 | 488 comments
A Thread Reader unrolled 37-tweet Twitter thread by Yam Peleg purporting to reveal leaked architecture and training details of OpenAI's GPT-4.
695 points by mfiguiere | 479 comments
OpenAI blog post announcing general availability of the GPT-4 API and the deprecation of older models in the Completions API.
322 points by georgehill | 167 comments
Tweet thread by swyx announcing a Latent Space podcast interview with George Hotz about tiny corp, including claims that GPT-4 is an 8x220B mixture model.
552 points by yonom | 232 comments
A Substack post by Simon Farshid exploring OpenAI's new function calling API for getting structured JSON output from GPT-4.
540 points by yla92 | 165 comments
OpenAI platform documentation guide outlining six strategies and tactics for getting better results from GPT models through prompting.
780 points by behnamoh | 631 comments
Ask HN thread where a user asks whether GPT-4's response quality has deteriorated recently, comparing ChatGPT, Phind, and Poe.
467 points by appwiz | 88 comments
GitHub repository from Brex containing a guide with tips and tricks for working with large language models like OpenAI's GPT-4.
467 points by mazzystar | 271 comments
A 2023 personal blog post reflecting on how six months of intensive ChatGPT/GPT-4 use gives individuals leverage to build things.
659 points by mfiguiere | 416 comments
OpenAI research post describing using GPT-4 to automatically generate and score explanations of individual neuron behavior in GPT-2.
1243 points by rushingcreek | 384 comments
Homepage of Phind, an AI-powered search engine for developers that answers technical questions, now offering a GPT-4 "Expert" mode.
380 points by adhocmobility | 742 comments
A personal essay on Notion lamenting GPT-4's impact on jobs, told through a colleague from a poor Indian farming family whose white-collar job may be automated.
397 points by obiefernandez | 207 comments
Obie Fernandez's Medium post recounting a Dungeons & Dragons session he played with his kids using ChatGPT-4 as the dungeon master.
949 points by jiripospisil | 887 comments
Future of Life Institute open letter calling on AI labs to pause training of AI systems more powerful than GPT-4 for at least six months, with a signature count.