State-of-the-Art Neurohive

LIMA: Pretraining Method on 1000 Examples Achieved GPT4-Level Accuracy

31 May 2023

LIMA: Pretraining Method on 1000 Examples Achieved GPT4-Level Accuracy

Language models typically undergo two stages of training: unsupervised pretraining and fine-tuning to specific tasks and user preferences. The novel LIMA method (Less Is More for Alignment) challenges the traditional…

Open-source model StarCoder generates code in 86 programming languages

10 May 2023

Open-source model StarCoder generates code in 86 programming languages

StarCoder is a state-of-the-art method for code correction and generation using neural networks from the research community The BigCode, MIT, University of Pennsylvania, and Columbia University. StarCoder improves quality and…

Google Imagen: text-to-image model

29 June 2022

Google Imagen: text-to-image model

Google has introduced Imagen, a model that transforms a text description into an image with a resolution of 1024 x 1024 pixels. Imagen surpassed OpenAI DALL-E 2 in terms of…

Deepmind has introduced a universal Gato model

28 May 2022

Deepmind has introduced a universal Gato model

DeepMind has introduced a cross-modal universal model with 1.2 billion Gato parameters. Gato can perform more than 600 tasks, such as playing video games, creating subtitles for images and controlling…

The model was trained to perform a cross-modal search for actions

9 May 2022

The model was trained to perform a cross-modal search for actions

MIT has developed a model of cross-modal search for actions in text, audio and video content. The model allows you to determine where a certain action takes place in the…

Flamingo: DeepMind multimodal model

9 May 2022

Flamingo: DeepMind multimodal model

Flamingo is a multimodal DeepMind model that generates a text description of photos, videos and sounds. The model surpasses the previous state-of-the-art models in 16 tasks, and its feature is…

MIT’s drone algorithm Predicts object Trajectories

29 April 2022

MIT’s drone algorithm Predicts object Trajectories

MIT researchers have developed an algorithm to improve the safety of self-driving cars. The model predicts the trajectories of road users moving near the drone in real time. Modern methods…

DALL-E 2: text-to-image OpenAI model

13 April 2022

DALL-E 2: text-to-image OpenAI model

OpenAI has introduced a new version of the DALL-E text-to-image conversion model. Compared to the first version, DALL-E 2 generates images in higher quality with less delay, and also allows…

PaLM: Google’s language model with 540 billion parameters

8 April 2022

PaLM: Google’s language model with 540 billion parameters

Google has introduced a PaLM – language model with 540 billion parameters. PaLM has surpassed existing language models in most benchmarks. The model is trained using 6144 Google TPU tensor…

The surgical robot determines the place of needle insertion

24 March 2022

The surgical robot determines the place of needle insertion

AI-Guide is a hand-held surgical robot developed at MIT that allows automating the process of inserting a needle or catheter into a blood vessel. The device is aimed at providing…

Reinforcement training to control thermonuclear reactions

17 February 2022

Reinforcement training to control thermonuclear reactions

DeepMind has announced the use of reinforcement learning to control the plasma state during a thermonuclear reaction. The DeepMind algorithm made it possible to increase the stability of the process…

Equidock: Prediction of protein complexes

8 February 2022

Equidock: Prediction of protein complexes

MIT has developed an Equidock neural network that predicts the connection of two proteins. The model can accelerate drug development by 500 times. Proteins produced by the immune system –…

Google AI has trained the robot to perform new tasks for it

5 February 2022

Google AI has trained the robot to perform new tasks for it

The Google AI study demonstrated the possibility of teaching robots to perform tasks that were not included in the training dataset. The method allows you to speed up and simplify…

The model was trained to find the optimal treatment regime

4 February 2022

The model was trained to find the optimal treatment regime

Microsoft has developed a reinforcement learning algorithm that offers the most effective treatment tactics for the patient’s current condition. The model is aimed at accelerating decision-making in healthcare in conditions…

OpenAI has trained a model to prove theorems

4 February 2022

OpenAI has trained a model to prove theorems

Open AIn AI presented a neural network proving theorems. The model achieved 41% accuracy on the miniF2F – dataset of school Olympiad problems. To search for evidence, a language model…

AlphaCode: code generation model as described by DeepMind

3 February 2022

AlphaCode: code generation model as described by DeepMind

DeepMind introduced the AlphaCode code generation system with 41 billion parameters. AlphaCode is superior to OpenAI Codex and generates code in 12 languages. According to a Cambridge University study, more…

The neural network has been trained to accurately separate the object from the background

25 January 2022

The neural network has been trained to accurately separate the object from the background

Google has developed a neural network that separates the object from the background in the image with high accuracy. The model is used in portrait shooting mode on Pixel 6.…

HyperStyle: photorealistic image editing

24 January 2022

HyperStyle: photorealistic image editing

HyperStyle is a neural network that modifies individual parameters of objects in photos. With HyperStyle, you can change a person’s hairstyle or the color of a car. A neural network…

A robot controlled by the patient’s brain

7 January 2022

A robot controlled by the patient’s brain

Researchers from the Federal Polytechnic School of Lausanne have developed a robot controlled on the basis of electrical signals coming from the brain. Such a robot can be used by…

GLIDE: Openair model for generating images by text

27 December 2021

GLIDE: Openair model for generating images by text

GLIDE is an OpenAI model for generating an image based on its description. GLIDE is superior to DALL-E and at the same time has 3 times fewer parameters. In January…

Openal trained the model to search for answers to questions on the Internet

19 December 2021

Openal trained the model to search for answers to questions on the Internet

OpenAI introduced WebGPT, a model that searches for the answer to a question on the Internet. WebGPT combines information from several sources and generates a response text. Language models such…