# Mar 21, 2024

### LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models

**LlamaFactory** introduces a **unified framework** for the efficient fine-tuning of over 100 large language models (LLMs), streamlining adaptation to various tasks without coding via LlamaBoard.

### Arcee's MergeKit: A Toolkit for Merging Large Language Models

**MergeKit** addresses the challenge of **combining the strengths of individual language models** into a single multitask model, enhancing both performance and versatility without the need for retraining.

### Parrots love playing tablet games. That's helping researchers understand them

**Parrots engage with tablet games** using their tongues and beaks, revealing unique interaction patterns that differ significantly from human touchscreen use, as observed in a study by Rébecca Kleinberger's lab at Northeastern University.

### Intel to Receive $8.5B in Grants to Build Chip Plants

**Intel has been awarded $8.5 billion in grants** by the U.S. government to support the construction and expansion of semiconductor plants in Arizona, Ohio, New Mexico, and Oregon, marking a significant step towards revitalizing domestic chip manufacturing.

### The baffling intelligence of a single cell: The story of E. coli chemotaxis

**E. coli chemotaxis** demonstrates a sophisticated form of intelligence, where the bacterium uses a complex signaling mechanism to navigate towards nutrients, showcasing a form of memory and decision-making without a brain.

### [D] Deanonymized paper accepted at ICLR 2024

A **paper** submitted to ICLR 2024 blatantly **violated double-blind review** protocols by including the first author's full name and acknowledgments, yet was initially reviewed without addressing this breach.

### So you think you want to write a deterministic hypervisor?

The **deterministic hypervisor**, dubbed **"the Determinator"**, is designed by Antithesis to emulate a deterministic computer, enabling the reproduction, exploration, and analysis of potential software bugs with unprecedented precision and reliability.

### Micrograd-CUDA: adapting Karpathy's tiny autodiff engine for GPU acceleration

**Micrograd-CUDA** is a project aimed at learning CUDA through the development of a GPU-accelerated, tensor-based automatic differentiation engine, inspired by [Andrej Karpathy's micrograd](https://github.com/karpathy/micrograd/tree/master), with no dependencies beyond Python's standard library and CUDA.

### Show HN: An AI-Powered WordPress Site Builder That We Are Open-Sourcing Today

**QuickWP**, an **AI-powered WordPress site builder**, leverages **OpenAI**, an **FSE theme**, and **WordPress Playground** to create personalized themes based on user input, now open-sourced for community learning and development.

### Launch HN: CamelQA (YC W24) – AI that tests mobile apps

**CamelQA** introduces an **AI agent** capable of automating mobile app testing through **natural language** and computer vision, targeting the reduction of time engineers spend on maintaining UI tests.

### Show HN: Personal Knowledge Base Visualization

**Knowledge** is a web application that aggregates and organizes content from social media platforms like GitHub, HackerNews, Zotero, and Twitter into a searchable knowledge graph, enhancing navigation and discovery.

### [D] Is there an accurate AI tool for research?

**AI tools like Perplexity, ChatGPT, and Bard** have shown variable accuracy in data analysis and providing up-to-date information, often requiring manual verification for reliability.

### Introducing pgzx: create PostgreSQL extensions using Zig

**pgzx is an open-source framework** for developing PostgreSQL extensions using Zig, offering utilities like error handling, memory allocators, and a development environment for easier integration with the Postgres codebase.

### [D] Why the readability of academic papers are continuously bad?

**Academic papers often assume extensive background knowledge** and use field-specific vocabulary, making them challenging to read without a solid foundation in the subject matter.

### [D] How much will Nvidia's newest Blackwell GPU's cut down training and inference time/price?

**Nvidia's newest Blackwell GPUs** are anticipated to significantly **reduce the training and inference costs** for models similar to Llama-2, which historically required approximately $5M.