Attention at Constant Cost per Token via Symmetry-Aware Taylor Approximation(arxiv.org)

158 points by fheinsen 8 months ago | 87 comments

Log-Linear Attention(arxiv.org)

19 points by sva_ 1 year ago | 3 comments

Multi-Token Attention(arxiv.org)

131 points by fzliu 1 year ago | 35 comments

HyperAttention: Long-Context Attention in Near-Linear Time(arxiv.org)

61 points by kelseyfrog 3 years ago | 10 comments

SplitQuantV2: Enhancing Low-Bit Quantization of LLMs Without GPUs(arxiv.org)

25 points by PaulHoule 1 year ago | 5 comments

QLoRA: Efficient Finetuning of Quantized LLMs(arxiv.org)

270 points by Garcia98 3 years ago | 90 comments

FSRS: A modern, efficient spaced repetition algorithm(github.com)

339 points by rickcarlino 2 years ago | 107 comments

GitHub wiki page for fsrs4anki explaining the basics of FSRS, a spaced repetition scheduling algorithm by Jarrett Ye intended to replace Anki's SM2.

GPT-5.2 derives a new result in theoretical physics(openai.com)

545 points by davidbarker 8 months ago | 369 comments

OpenAI post announcing that GPT-5.2 derived a new result in theoretical physics; the captured page shows only a loading logo.

What AI-Native Looks Like(fagnerbrack.com)

14 points by fagnerbrack 1 month ago | 7 comments

A blog post by Fagner Brack on Medium describing what an AI-native approach to software development with stable infrastructure looks like.

Training language models to be warm and empathetic makes them less reliable(arxiv.org)

334 points by Cynddl 1 year ago | 333 comments

arXiv paper finding that fine-tuning language models to be warm and empathetic raises error rates and makes them more sycophantic on safety-critical tasks.

The Era of 1-bit LLMs: ternary parameters for cost-effective computing(arxiv.org)

962 points by fgfm 2 years ago | 419 comments

arXiv paper introducing BitNet b1.58, a ternary-weight (-1, 0, 1) LLM matching full-precision Transformers while being far cheaper in latency, memory, and energy.

An Overwhelmingly Negative and Demoralizing Force(aftermath.site)

394 points by Doches 1 year ago | 362 comments

An Aftermath article about how generative AI tools like Midjourney and vibe coding are affecting video game developers and artists, described as a demoralizing force.