I will create genuine hard terminal bench and swe bench benchmark tasks

India

I speak Hindi, English

AI evaluation

I am a B.Tech. candidate with hands-on experience creating and reviewing Terminal bench and SWE bench benchmark tasks for frontier AI agents and building Python ML/RAG systems. I specialize in AI agen...
About this Gig

I am an experienced Terminal Bench and SWE bench benchmark task author.


I have created and submitted 32 hard Terminal-Bench 2.1 tasks through Snorkel AI platform for frontier AI agent training and evaluation out of which 21 submissions were accepted.


I have been selected as a reviewer for reviewal of SWE-bench benchmark tasks to support quality and task-readiness review. I reviewed 83 submitted SWE-bench tasks and implemented fixes where applicable.


I have also submitted 21 SWE bench benchmark tasks, and have 17 accepted out of them


I have also created of 13 tasks of Terminal Bench 3.0 so far.

Number of tasks created of 18 tasks of Terminal Bench 4.0 so far.


In this gig I'm offering tasks of the following benchmarks:


  1. Terminal Bench 2.1
  2. Terminal Bench 3.0
  3. Terminal Bench 4.0
  4. SWE Bench


There are three packages(Basic, Standard, and Premium) containing different number of tasks in each package.


I guarantee high-quality and high genuine difficulty(avoiding False Failures) tasks irrespective of the package you buy.

Related tags