BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Ainave//Events//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VEVENT
UID:dorahacks.io/hackathon/binance/detail
DTSTAMP:20261006T001929Z
DTSTART;VALUE=DATE:20260918
DTEND;VALUE=DATE:20261114
SUMMARY:Beyond the Benchmark: The Binance Agentic AI Challenge
DESCRIPTION:### Overview\nBenchmarks are static. Reality isn’t. This Binance Agentic AI Challenge asks teams to build an LLM powered\, autonomous agent that can gather information\, reason\, decide\, manage its own risk\, and operate reliably in a dynamic\, real time environment. The competition runs online with a multi stage process designed to test end to end autonomy and reliability\, not just theoretical design. Applications for proposals are open and guidance is provided to finalists through onboarding materials\, audits\, and case documentation.\n\n### The Challenge\nParticipants must create an autonomous Agentic AI system capable of operating in the real world. Unlike typical simulations\, this challenge evaluates performance during a 14 day window with real time conditions\, latency\, and cost constraints. The system should acquire data\, reason through problems\, use tools\, manage risk\, and execute eligible trades on the Binance Smart Chain via Binance Web3 Open APIs.\n\nYou build\, host\, and operate the system while Binance sets the rules and conducts the audit for integrity. Teams own their wallets\, keys\, data pipelines\, models\, and hosting\, while Binance owns the rules\, instruments\, the scored window\, the audit standard\, and the final results. The repository is private and audited by a Binance account for integrity review.\n\n### What makes a strong entry\nEntries must be genuinely agentic. Excellence occurs at both the model layer (domain post training\, fine tuning\, tool use\, memory\, planning\, verification) and the harness layer (context\, tools\, memory\, guardrails\, observability\, recovery). Data pipelines are considered a competitive edge\, covering data sources\, services\, and hosting costs. Reliability matters\; failures and recovery are part of the audit evidence. Every scored trade should be traceable to a registered Agent version and a unique decision record. The rules are the same for all teams\, with onboarding providing full technical parameters.\n\n### Timeline & Submission\nApplications include an initial proposal phase\, followed by a 14 day scored run and a winner audit. Finalists are selected from proposals\, and there is a separate stage for the scored run where on chain state and execution records determine the final score. There is no public real-time leaderboard\; the entire 14 day window counts toward the score. Finalists receive a Participant Guide\, API documentation\, an Audit Log Specification\, scoring rules\, and incident policy at onboarding. The page outlines ongoing submission guidelines and the requirement to demonstrate end-to-end operation during the scored period.\n\n### Judging & Scoring\nTwo separate evaluations occur: proposals are judged by people and the ranking is not. Stage 1 selects finalists by feasibility\, agentic depth\, risk awareness\, originality\, and ability to deliver. Stage 2 is the 14 day mechanical scored run. The score is based on wallet equity changes\, on chain state\, and the final score formula. Stage 3 is a winner audit to verify the integrity of the results. Not scored are presentation polish and other non technical factors. The final scoring method\, transfers\, and tie-breakers are confirmed in the onboarding rules.\n\n### Rules & Restrictions\nAutonomous operation is mandatory\; every trade must originate from the registered Agentic AI system. Repository integrity and logging rules prohibit secret or private data exposure and require proper versioning and manifests. Honest logging and clean information rules ensure traceability and auditable decisions. Dedicated wallets that involve external funds\, unapproved transfers\, or manual trading are prohibited. The audit can remove a result if irregularities are found. Eligibility is open to individuals who meet local law requirements\; entities\, officials with restricted access\, and Binance employees are excluded. Onboarding provides final operating parameters to finalists.\n\n### Eligibility\nOpen to undergraduate and graduate students\, professionals\, and academic teams applying in their personal capacity. Participants must own all data sources\, models\, and components and complete identity verification during onboarding. Binance may verify eligibility at any stage. This is a technical research and developer challenge rather than trading education or investment advice.\n\n### Audience\nThe event targets AI and automation enthusiasts\, blockchain developers\, and teams seeking to showcase autonomous agentic capabilities in real time. It emphasizes end-to-end operation\, auditability\, and accountable decision making within a structured\, rules-based competition framework.\nhttps://dorahacks.io/hackathon/binance/detail?ref=ainave
LOCATION:
URL;VALUE=URI:https://dorahacks.io/hackathon/binance/detail?ref=ainave
END:VEVENT
END:VCALENDAR