Primary competition visual

TelcoRetrieve Challenge: Unlocking Telco Knowledge by ITU

$20 000 USD
Starting soon! (1 day to launch)
Retrieval-Augmented Generation (RAG)
Information Retrieval
NLP
Model evaluation & Benchmarking
LLM fine-tuning
92 joined
0 active
Starti
13 Nov 26
Enrolments closei
06 Jan 27
Closei
13 Jan 27
Reveali
31 Jan 27
About

Indexed data and QnA: telco-retrieve-challenge - a GSMA Collection

TelcoRetrieve Pipeline: Znbne/telco-retrieve: rag pipeline for telecom specs

Training dataset: telecom-standards-corpus - a GSMA Collection

Raw 3GPP dataset: GSMA/3GPP · Datasets at Hugging Face

Constraints

  • Generator Model: Gemma 3n E4B instruct (google/gemma-3n-e4b-it)​
  • Embedding Model: farbodtavakkoli/OTel-Embedding-300M
  • Reranker Model: farbodtavakkoli/OTel-Reranker-0.6B
  • Accuracy Min Threshold: 80% (the submission with accuracy below that will be disqualified)

Available AMD GPU Configuration (2 Instances Available)

We will have two instances available from 15 Jan 2027, each with the following specifications:

  • Type: Virtual Machine / Cloud Instance
  • VRAM: 288 GB HBM3E
  • vCPU: 24 to 32 vCPUs
  • Boot Disk: 720 GB NVMe SSD
  • NVMe SSD Scratch Disk: 5 TB local NVMe SSD

Files
Description
Files