Datasets:

gradients-io-tournaments
/

bittensor-subnets-qa

id stringlengths 7 19	category stringlengths 3 16	conversations listlengths 2 2	question stringlengths 17 65	answer stringlengths 442 806
bittensor_001	fundamentals	[ { "content": "What is Bittensor?", "role": "user" }, { "content": "Bittensor is a decentralized blockchain network designed to enhance the collaboration of artificial intelligence (AI) and machine learning models within a peer-to-peer ecosystem. Created in 2021 by the Opentensor Foundation (founded ...	What is Bittensor?	Bittensor is a decentralized blockchain network designed to enhance the collaboration of artificial intelligence (AI) and machine learning models within a peer-to-peer ecosystem. Created in 2021 by the Opentensor Foundation (founded by Jacob Steeves and Ala Shaabana), Bittensor creates an open and decentralized marketp...
bittensor_002	fundamentals	[ { "content": "How does Bittensor work?", "role": "user" }, { "content": "Bittensor operates as a decentralized network of AI models organized into specialized sub-networks called subnets. Each subnet focuses on specific AI tasks where miners run AI models to compete based on specific criteria. The n...	How does Bittensor work?	Bittensor operates as a decentralized network of AI models organized into specialized sub-networks called subnets. Each subnet focuses on specific AI tasks where miners run AI models to compete based on specific criteria. The network consists of three key elements: 1) The foundational blockchain (Subtensor) which opera...
bittensor_003	fundamentals	[ { "content": "What is the TAO token and what is it used for?", "role": "user" }, { "content": "TAO is Bittensor's native cryptocurrency token with multiple roles in the ecosystem. It serves as: 1) A utility token for accessing AI services on the network, 2) A staking token for network security where...	What is the TAO token and what is it used for?	TAO is Bittensor's native cryptocurrency token with multiple roles in the ecosystem. It serves as: 1) A utility token for accessing AI services on the network, 2) A staking token for network security where validators stake TAO to secure the network, 3) Payment for transaction fees when registering AI models or submitti...
bittensor_004	fundamentals	[ { "content": "What is Proof of Intelligence in Bittensor?", "role": "user" }, { "content": "Proof of Intelligence (PoI), also known as Yuma Consensus, is Bittensor's unique consensus mechanism. Unlike traditional Proof of Work (solving mathematical problems) or Proof of Stake (staking tokens), PoI r...	What is Proof of Intelligence in Bittensor?	Proof of Intelligence (PoI), also known as Yuma Consensus, is Bittensor's unique consensus mechanism. Unlike traditional Proof of Work (solving mathematical problems) or Proof of Stake (staking tokens), PoI rewards participants for the value of their machine learning contributions. Nodes are required to perform machine...
bittensor_005	fundamentals	[ { "content": "What are subnets in Bittensor?", "role": "user" }, { "content": "Subnets are specialized sub-networks within Bittensor, each with its own unique algorithm and incentives. They leverage Bittensor's infrastructure to build and run AI-powered projects. Think of them as independent markets...	What are subnets in Bittensor?	Subnets are specialized sub-networks within Bittensor, each with its own unique algorithm and incentives. They leverage Bittensor's infrastructure to build and run AI-powered projects. Think of them as independent markets where participants contribute computing power to perform specific tasks like text generation, imag...
yuma_001	consensus	[ { "content": "How does Yuma Consensus work?", "role": "user" }, { "content": "Yuma Consensus (YC) is the critical algorithmic process running on-chain within Subtensor that computes validator and miner emissions from validators' rankings. Each validator periodically submits a vector of weights ranki...	How does Yuma Consensus work?	Yuma Consensus (YC) is the critical algorithmic process running on-chain within Subtensor that computes validator and miner emissions from validators' rankings. Each validator periodically submits a vector of weights ranking miners' performance. The algorithm resolves this matrix into two emissions vectors—one for mine...
yuma_002	consensus	[ { "content": "What is clipping in Yuma Consensus?", "role": "user" }, { "content": "Clipping is a mechanism in Yuma Consensus designed to punish inaccurate miner evaluation and prevent collusion. It works by using the judgment of the most trusted validators (measured by stake) as a benchmark. If the...	What is clipping in Yuma Consensus?	Clipping is a mechanism in Yuma Consensus designed to punish inaccurate miner evaluation and prevent collusion. It works by using the judgment of the most trusted validators (measured by stake) as a benchmark. If the least generous half of validators (by stake) set weights for a miner to no more than x, then the more g...
yuma_003	consensus	[ { "content": "How are rewards distributed in Bittensor?", "role": "user" }, { "content": "Bittensor distributes rewards through Yuma Consensus. Of total emissions: 41% goes to miners (allocated based on consensus scores), 41% goes to validators (with 82% of validator rewards going to TAO stakers), a...	How are rewards distributed in Bittensor?	Bittensor distributes rewards through Yuma Consensus. Of total emissions: 41% goes to miners (allocated based on consensus scores), 41% goes to validators (with 82% of validator rewards going to TAO stakers), and 18% goes to subnet owners. The better a miner is rated, the more they're paid. The closer a validator's rat...
dtao_001	dynamic_tao	[ { "content": "What is Dynamic TAO (dTAO)?", "role": "user" }, { "content": "Dynamic TAO (dTAO) is a major evolution of Bittensor's tokenomics and governance model, deployed on mainnet on February 13, 2025. It fundamentally changes how the network allocates TAO tokens. Previously, a small group of 'r...	What is Dynamic TAO (dTAO)?	Dynamic TAO (dTAO) is a major evolution of Bittensor's tokenomics and governance model, deployed on mainnet on February 13, 2025. It fundamentally changes how the network allocates TAO tokens. Previously, a small group of 'root' validators had outsized influence over emissions, leading to potential cronyism or collusio...
dtao_002	dynamic_tao	[ { "content": "What are Alpha tokens in Bittensor?", "role": "user" }, { "content": "Alpha tokens are subnet-specific tokens introduced with Dynamic TAO. Each subnet has its own alpha token (e.g., ALPHA for subnet 1, BETA for subnet 2) capped at 21 million with a halving schedule like TAO. When you s...	What are Alpha tokens in Bittensor?	Alpha tokens are subnet-specific tokens introduced with Dynamic TAO. Each subnet has its own alpha token (e.g., ALPHA for subnet 1, BETA for subnet 2) capped at 21 million with a halving schedule like TAO. When you stake TAO, you choose which subnet to stake to, and your TAO gets exchanged for that subnet's alpha token...
dtao_003	dynamic_tao	[ { "content": "How has Dynamic TAO impacted Bittensor?", "role": "user" }, { "content": "Dynamic TAO has dramatically transformed Bittensor. Over 100 subnets have launched since dTAO's debut, covering diverse AI tasks from language model inference to GPU compute and data pipelines. Staking has become...	How has Dynamic TAO impacted Bittensor?	Dynamic TAO has dramatically transformed Bittensor. Over 100 subnets have launched since dTAO's debut, covering diverse AI tasks from language model inference to GPU compute and data pipelines. Staking has become a speculative position—if demand for a subnet rises, its alpha token price increases; if popularity falls, ...
gradients_001	gradients	[ { "content": "What is Gradients subnet on Bittensor?", "role": "user" }, { "content": "Gradients is Bittensor Subnet 56, developed by Rayon Labs and owned by Grads LLC. Its tagline is 'Anyone Can Train on Bittensor - Gradients on Demand.' It's the first and only platform enabling decentralized model...	What is Gradients subnet on Bittensor?	Gradients is Bittensor Subnet 56, developed by Rayon Labs and owned by Grads LLC. Its tagline is 'Anyone Can Train on Bittensor - Gradients on Demand.' It's the first and only platform enabling decentralized model training on Bittensor, addressing the previously under-served need for fine-tuning AI models. Gradients ha...
gradients_002	gradients	[ { "content": "How does Gradients subnet work?", "role": "user" }, { "content": "Gradients allows users to easily launch fine-tuning sessions without managing heavy infrastructure. All technical aspects—from orchestration to GPU resource management—are handled by Subnet 56 miners. Users can upload a ...	How does Gradients subnet work?	Gradients allows users to easily launch fine-tuning sessions without managing heavy infrastructure. All technical aspects—from orchestration to GPU resource management—are handled by Subnet 56 miners. Users can upload a dataset, select a model, and have Bittensor miners compete to produce the best-performing version, r...
gradients_003	gradients	[ { "content": "What can I train on Gradients?", "role": "user" }, { "content": "Gradients enables training of any small- to medium-sized LLM from Hugging Face. The platform offers a wide variety of models and datasets. With just a few clicks, anyone regardless of AI training knowledge can train model...	What can I train on Gradients?	Gradients enables training of any small- to medium-sized LLM from Hugging Face. The platform offers a wide variety of models and datasets. With just a few clicks, anyone regardless of AI training knowledge can train models. The platform provides intelligent AI tools that automatically optimize your training process. Us...
gradients_004	gradients	[ { "content": "How much does Gradients cost?", "role": "user" }, { "content": "Gradients charges $5 per hour of training, which is significantly cheaper than competitors. It's cheaper than Together AI and far cheaper than major cloud solutions from Amazon and Google. The platform has attracted over 3...	How much does Gradients cost?	Gradients charges $5 per hour of training, which is significantly cheaper than competitors. It's cheaper than Together AI and far cheaper than major cloud solutions from Amazon and Google. The platform has attracted over 3,000 paying users, primarily hobbyists, and delivers results that are both cheaper and better than...
gradients_005	gradients	[ { "content": "Who created Gradients subnet?", "role": "user" }, { "content": "Gradients was created by Rayon Labs, a leading developer team on Bittensor. The team is led by 'wanderingweights' who created this no-code training subnet. Rayon Labs currently operates three major subnets: Chutes (Subnet ...	Who created Gradients subnet?	Gradients was created by Rayon Labs, a leading developer team on Bittensor. The team is led by 'wanderingweights' who created this no-code training subnet. Rayon Labs currently operates three major subnets: Chutes (Subnet 64), Gradients (Subnet 56), and Nineteen (Subnet 19), which rank among the most capitalized in the...
gradients_006	gradients	[ { "content": "What are the latest updates to Gradients?", "role": "user" }, { "content": "Gradients V3 was launched on April 15, 2025, bringing significant improvements. The subnet has seen remarkable growth, with its token up over 500% in just a couple of weeks in early April 2025. Key expansions i...	What are the latest updates to Gradients?	Gradients V3 was launched on April 15, 2025, bringing significant improvements. The subnet has seen remarkable growth, with its token up over 500% in just a couple of weeks in early April 2025. Key expansions include: 1) Full-featured image model training tools with specialized infrastructure optimized for efficient im...
gradients_007	gradients	[ { "content": "How do I use Gradients to train a model?", "role": "user" }, { "content": "Using Gradients is simple: 1) Access the platform through the Gradients UI or programmatically via their API, 2) Select from the wide variety of available models from Hugging Face, 3) Upload your custom dataset ...	How do I use Gradients to train a model?	Using Gradients is simple: 1) Access the platform through the Gradients UI or programmatically via their API, 2) Select from the wide variety of available models from Hugging Face, 3) Upload your custom dataset or choose from available datasets, 4) Configure your training preferences (the platform's intelligent tools w...
gradients_008	gradients	[ { "content": "What emissions does Gradients subnet receive?", "role": "user" }, { "content": "Gradients (Subnet 56) pulls in approximately 6.4% of all TAO emissions, which translates to about 460 TAO daily. This breaks down to: 83 TAO for subnet owners and 189 TAO each for miners and validators. Thi...	What emissions does Gradients subnet receive?	Gradients (Subnet 56) pulls in approximately 6.4% of all TAO emissions, which translates to about 460 TAO daily. This breaks down to: 83 TAO for subnet owners and 189 TAO each for miners and validators. This substantial emission share reflects the high demand and utility of the training services provided. Combined with...
chutes_001	chutes	[ { "content": "What is Chutes subnet on Bittensor?", "role": "user" }, { "content": "Chutes (Subnet 64) is a decentralized serverless AI compute platform built on Bittensor, developed by Rayon Labs. It provides an open, on-demand AI inference service that lets developers deploy, run, and scale AI mod...	What is Chutes subnet on Bittensor?	Chutes (Subnet 64) is a decentralized serverless AI compute platform built on Bittensor, developed by Rayon Labs. It provides an open, on-demand AI inference service that lets developers deploy, run, and scale AI models in seconds without managing any infrastructure. Developers can instantly deploy, monetize, and scale...
chutes_002	chutes	[ { "content": "How does Chutes work?", "role": "user" }, { "content": "In Chutes, miners work by running packages called 'chutes'—essentially code that can be AI models, data processing applications, or any computational task. Developers deploy a chute via API or CLI, specifying requirements like GPU...	How does Chutes work?	In Chutes, miners work by running packages called 'chutes'—essentially code that can be AI models, data processing applications, or any computational task. Developers deploy a chute via API or CLI, specifying requirements like GPU count and memory. The Chutes scheduler assigns the model to suitable miner nodes meeting ...
chutes_003	chutes	[ { "content": "What are the benefits of using Chutes?", "role": "user" }, { "content": "Chutes offers several key benefits: 1) Costs up to 90% lower than specialized platforms like major cloud providers, 2) Support for a wide range of model types from LLMs to image and audio models, 3) Distributed in...	What are the benefits of using Chutes?	Chutes offers several key benefits: 1) Costs up to 90% lower than specialized platforms like major cloud providers, 2) Support for a wide range of model types from LLMs to image and audio models, 3) Distributed infrastructure reducing single-point-of-failure risks, 4) Automatic scaling with subnet demand, 5) Anti-sybil...
bitmind_001	bitmind	[ { "content": "What is BitMind subnet on Bittensor?", "role": "user" }, { "content": "BitMind is Bittensor Subnet 34, focused on deepfake detection and AI-generated content verification. Its mission is to create the best deepfake detection solution on the market, demonstrating the potential of decent...	What is BitMind subnet on Bittensor?	BitMind is Bittensor Subnet 34, focused on deepfake detection and AI-generated content verification. Its mission is to create the best deepfake detection solution on the market, demonstrating the potential of decentralized AI. BitMind incentivizes identifying and mitigating AI-generated image content, achieving up to 9...
bitmind_002	bitmind	[ { "content": "How does BitMind detect deepfakes?", "role": "user" }, { "content": "BitMind uses a multi-model framework called Content-Aware Model Orchestration (CAMO). Miners receive images and output binary classifications—1 for deepfake, 0 for real. CAMO uses a gating mechanism for model orchestr...	How does BitMind detect deepfakes?	BitMind uses a multi-model framework called Content-Aware Model Orchestration (CAMO). Miners receive images and output binary classifications—1 for deepfake, 0 for real. CAMO uses a gating mechanism for model orchestration, dynamically routing content-specific images to appropriate detectors. BitMind also open-sourced ...
bitmind_003	bitmind	[ { "content": "What is GAS in BitMind?", "role": "user" }, { "content": "GAS (Generative Adversarial Subnet) is BitMind's adversarial training system inspired by GANs. Detectors and generators compete in a dynamic loop: detectors sharpen their ability to spot synthetic media while generators push to ...	What is GAS in BitMind?	GAS (Generative Adversarial Subnet) is BitMind's adversarial training system inspired by GANs. Detectors and generators compete in a dynamic loop: detectors sharpen their ability to spot synthetic media while generators push to create more convincing fakes. This adversarial process drives cutting-edge detection tools a...
bitmind_004	bitmind	[ { "content": "What products does BitMind offer?", "role": "user" }, { "content": "BitMind offers several consumer products: 1) Discord and Telegram bots allowing users to classify images as real or AI-generated directly in messaging apps, 2) Agent Smith browser extension that examines webpage images...	What products does BitMind offer?	BitMind offers several consumer products: 1) Discord and Telegram bots allowing users to classify images as real or AI-generated directly in messaging apps, 2) Agent Smith browser extension that examines webpage images in real-time, querying subnet validators and producing real vs. fake verification overlays on visible...
nineteen_001	nineteen	[ { "content": "What is Subnet 19 Vision on Bittensor?", "role": "user" }, { "content": "Subnet 19, also known as Nineteen or Vision, is a Bittensor subnetwork focused on inference at scale, with initial focus on image generation and manipulation models. Developed by Namoray, one of Bittensor's premie...	What is Subnet 19 Vision on Bittensor?	Subnet 19, also known as Nineteen or Vision, is a Bittensor subnetwork focused on inference at scale, with initial focus on image generation and manipulation models. Developed by Namoray, one of Bittensor's premier developers (and part of Rayon Labs), it empowers Bittensor with image recognition using Meta's Segment An...
nineteen_002	nineteen	[ { "content": "What capabilities does Vision subnet have?", "role": "user" }, { "content": "Vision subnet has evolved significantly: Image recognition with Segment Anything (object detection, segmentation, embeddings), CLIP models for image and text search, Text-to-Image and Image-to-Image generation...	What capabilities does Vision subnet have?	Vision subnet has evolved significantly: Image recognition with Segment Anything (object detection, segmentation, embeddings), CLIP models for image and text search, Text-to-Image and Image-to-Image generation, Upscaler and Inpainting models, Decentralized Subnet Inference at Scale (DSIS) framework, MODE (Mixture Of De...
nineteen_003	nineteen	[ { "content": "How does Vision subnet's DSIS work?", "role": "user" }, { "content": "Decentralized Subnet Inference at Scale (DSIS) is Vision's framework to maximize output capacity of the Bittensor network. It allows miners to freely choose their technology stack to process requirements and generate...	How does Vision subnet's DSIS work?	Decentralized Subnet Inference at Scale (DSIS) is Vision's framework to maximize output capacity of the Bittensor network. It allows miners to freely choose their technology stack to process requirements and generate responses. Validators receive requirements from the front end, distribute them to miners, evaluate perf...
sportstensor_001	sportstensor	[ { "content": "What is Sportstensor on Bittensor?", "role": "user" }, { "content": "Sportstensor is Bittensor Subnet 41, a decentralized sports prediction subnet. It leverages the collective intelligence of miners to generate accurate forecasts for sporting events. The subnet is designed to incentivi...	What is Sportstensor on Bittensor?	Sportstensor is Bittensor Subnet 41, a decentralized sports prediction subnet. It leverages the collective intelligence of miners to generate accurate forecasts for sporting events. The subnet is designed to incentivize discovering competitive advantages ('edge') over closing market odds, enabling top miners to establi...
sportstensor_002	sportstensor	[ { "content": "How does Sportstensor work with Polymarket?", "role": "user" }, { "content": "Sportstensor has a product called Almanac—a prediction-market layer built on top of Polymarket. The system aggregates predictions from miners into a meta-model and trades with its own bankroll on Polymarket. ...	How does Sportstensor work with Polymarket?	Sportstensor has a product called Almanac—a prediction-market layer built on top of Polymarket. The system aggregates predictions from miners into a meta-model and trades with its own bankroll on Polymarket. Users can also route their own bets through Almanac. The platform charges a 1% fee on winning trades, but replac...
sportstensor_003	sportstensor	[ { "content": "How are Sportstensor miners evaluated?", "role": "user" }, { "content": "Miners run ML models analyzing historical and real-time data to predict outcomes. Validators send requests every 15 minutes to check miners' committed leagues. Miners must respond with predictions for all matches ...	How are Sportstensor miners evaluated?	Miners run ML models analyzing historical and real-time data to predict outcomes. Validators send requests every 15 minutes to check miners' committed leagues. Miners must respond with predictions for all matches in their committed league or receive penalties. The scoring compares predictions against closing odds—the p...
pretraining_001	pretraining	[ { "content": "What is the Pretraining subnet on Bittensor?", "role": "user" }, { "content": "Subnet 9 is Bittensor's premier training subnet, rewarding miners for producing pretrained Foundation Models on the Falcon Refined Web dataset. It acts as a continuous benchmark where miners are rewarded for...	What is the Pretraining subnet on Bittensor?	Subnet 9 is Bittensor's premier training subnet, rewarding miners for producing pretrained Foundation Models on the Falcon Refined Web dataset. It acts as a continuous benchmark where miners are rewarded for attaining the best losses on randomly sampled pages of Falcon given a consistent model architecture. The subnet ...
pretraining_002	pretraining	[ { "content": "How does Subnet 9 Pretraining work?", "role": "user" }, { "content": "The mechanism works in three steps: 1) Miners train and periodically publish models to Hugging Face, committing metadata to the Bittensor chain, 2) Validators download models based on chain metadata and continuously ...	How does Subnet 9 Pretraining work?	The mechanism works in three steps: 1) Miners train and periodically publish models to Hugging Face, committing metadata to the Bittensor chain, 2) Validators download models based on chain metadata and continuously evaluate them, setting weights based on performance against the Falcon dataset, 3) The chain aggregates ...
pretraining_003	pretraining	[ { "content": "What models has Subnet 9 produced?", "role": "user" }, { "content": "Subnet 9 has produced competitive open-source models. Notably, Tensorplex Labs published Sumo-T9-7B, a top-performing model that surpassed notable models like Falcon 7B and Llama 2 7B within less than a month of the 7...	What models has Subnet 9 produced?	Subnet 9 has produced competitive open-source models. Notably, Tensorplex Labs published Sumo-T9-7B, a top-performing model that surpassed notable models like Falcon 7B and Llama 2 7B within less than a month of the 7B parameter limit upgrade. The subnet's design prevents miners or validators from cheating, creating a ...
ptn_001	ptn	[ { "content": "What is the Proprietary Trading Network on Bittensor?", "role": "user" }, { "content": "The Proprietary Trading Network (PTN) is Subnet 8, developed by Taoshi. Initially launched in November 2023 as a time-series prediction subnet for forecasting Bitcoin prices, it pivoted to a sophist...	What is the Proprietary Trading Network on Bittensor?	The Proprietary Trading Network (PTN) is Subnet 8, developed by Taoshi. Initially launched in November 2023 as a time-series prediction subnet for forecasting Bitcoin prices, it pivoted to a sophisticated trading system focused on generating and executing risk-adjusted strategies for cryptocurrency and forex markets. T...
ptn_002	ptn	[ { "content": "How does Taoshi's prediction subnet work?", "role": "user" }, { "content": "For financial markets, miners predict the next 8 hours of closes on 5-minute intervals (100 closes). At minimum, miners should use OHLCV data plus additional sources to be competitive. Miners are rewarded based...	How does Taoshi's prediction subnet work?	For financial markets, miners predict the next 8 hours of closes on 5-minute intervals (100 closes). At minimum, miners should use OHLCV data plus additional sources to be competitive. Miners are rewarded based on predicting live data that hasn't occurred yet—not historical data. The subnet has expanded beyond Bitcoin ...
filetao_001	filetao	[ { "content": "What is FileTAO on Bittensor?", "role": "user" }, { "content": "FileTAO is Subnet 21, a decentralized distributed storage subnet on Bittensor. It represents a significant advancement in decentralized storage, providing an efficient, secure, and private solution. FileTAO implements a no...	What is FileTAO on Bittensor?	FileTAO is Subnet 21, a decentralized distributed storage subnet on Bittensor. It represents a significant advancement in decentralized storage, providing an efficient, secure, and private solution. FileTAO implements a novel multi-layered zero-knowledge interactive proof-of-spacetime algorithm using Pedersen commitmen...
filetao_002	filetao	[ { "content": "How does FileTAO ensure privacy?", "role": "user" }, { "content": "FileTAO validates on encrypted user data, meaning miners are unaware of what data they're storing. Only end-users can encrypt/decrypt their data using their Bittensor wallet coldkey. The system uses Content Identifiers ...	How does FileTAO ensure privacy?	FileTAO validates on encrypted user data, meaning miners are unaware of what data they're storing. Only end-users can encrypt/decrypt their data using their Bittensor wallet coldkey. The system uses Content Identifiers (CIDs) for content addressing rather than location-based addressing, which is inherently more secure ...
apex_001	apex	[ { "content": "What is the Apex/Text Prompting subnet on Bittensor?", "role": "user" }, { "content": "Subnet 1 (Apex) is Bittensor's flagship subnet focused on incentivizing conversational intelligence growth globally. As the inaugural Bittensor subnet dedicated to text generation, it supports prompt...	What is the Apex/Text Prompting subnet on Bittensor?	Subnet 1 (Apex) is Bittensor's flagship subnet focused on incentivizing conversational intelligence growth globally. As the inaugural Bittensor subnet dedicated to text generation, it supports prompt-based neural networks like GPT-3, GPT-4, and ChatGPT in a decentralized fashion. Both validators and miners use Large La...