Sentient launches the latest season of Hackathon Challenge 0, offering over $6,000 and points rewards

robot
Abstract generation in progress
Mars Finance News, Sentient announced that from May 20 to June 22, it will host the latest season of hackathon competition "Challenge 0," with a total prize pool including over $6,000 and MiniMax points. It is reported that this challenge focuses on reasoning based on real financial data, using Databricks' OfficeQA benchmark test. Participating teams are required to build an AI Agent that reads and reasons financial tables based on historical documents from the U.S. Department of the Treasury (1939-2025) to answer 246 financial questions. The official states that developers do not need to write code and can participate directly using pre-built coding Agents, optimizing through prompt engineering, MCP servers, and other methods. The system will evaluate and rank based on accuracy, latency, and cost.
SENT2.62%
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • 9
  • 12
  • Share
Comment
Add a comment
Add a comment
AirdropLunchbox
· 11h ago
Accuracy + latency + cost triangle balance, this evaluation criterion is much more interesting than just boosting rankings.
View OriginalReply0
GateUser-715706bb
· 05-21 02:36
Historical financial table reasoning, this application scenario is more practical than imagined, and can be used for audit compliance as well.
View OriginalReply0
BreadthHunter
· 05-21 02:31
246 questions, it feels like in the end it will turn into a competition of who’s API calls are more cost-effective.
View OriginalReply0
GlassDomeUniverse
· 05-21 02:30
Treasury Department documents from 1939? This dataset spans quite a range; an agent would probably go crazy trying to understand handwritten tables from 80 years ago.
View OriginalReply0
GasFeesForNightRuns
· 05-21 02:24
Pre-constructed encoding Agent + prompt engineering, the low-code track is heating up.
View OriginalReply0
ExitLiquidityStan
· 05-21 02:24
Databricks OfficeQA benchmark, evaluation dimensions also consider latency and cost, very realistic
View OriginalReply0
MoonlightDisconnectSwitch
· 05-21 02:24
Does anyone know about MCP server optimization? Looking for a configuration tutorial.
View OriginalReply0
NonceNinja
· 05-21 02:24
Are you serious about not needing programming? So finally, someone like me who only knows how to write prompts can get a seat at the table.
View OriginalReply0
OrdersPlacedBeforeTheStorm
· 05-21 02:24
$6,000 prize pool + MiniMax points, rounded to the nearest whole number, equals free mining power and even a profit, so I went all in.
View OriginalReply0
View More
  • Pinned