# We use the ir-metadata standard to describe and process runs.
# The fields below are mandatory, you can add additional metadata if you like.
# There are two libraries that automate and simplify tracking of experimental metadata that you can use optionally:
#
#   - The metadata module of [repro_eval](https://www.ir-metadata.org/software/):
#     Tracks the platform and implementation of your experiments.
#
#   - The [tirex_tracker](https://github.com/tira-io/tirex-tracker/):
#     Tracks the platform, implementation, and resources (e.g., CPU, GPU, RAM, emissions, etc.) of your experiments.
#
# See https://www.ir-metadata.org for more details.

tag: ENTER_VALUE_HERE

actor:
  # The name of the team
  team: ENTER_VALUE_HERE

# Please provide a short description of your system.
research goal:
  description: |
    ENTER_VALUE_HERE
    ...
    This is a multi line string
    ...
    ENTER_VALUE_HERE

platform:
  software:
    # Which software and tools did you use for training, tunning and running your system?
    # You can maintain the software that you used manually.
    # Alternatively, you can use repro_eval or the tirex_tracker to track this.
    libraries:
      - ENTER_VALUE_HERE
      - ENTER_VALUE_HERE

implementation:
  source:
    # Please provide a reference to your code if possible.
    # If you can not share your code, you can delete the implementation section.
    # The repro_eval and/or tirex_tracker tools can track this automatically, including commits, branches, etc.
    repository: ENTER_VALUE_HERE

data:
  # Please describe which training data your system used, e.g., longeval-sci-2026, longeval-web, MS MARCO, etc.
  training data:
    - name: longeval-sci-2026
    - name: ENTER_VALUE_HERE

method:
  # Boolean value indicating if it is a automatic (true) or manual (false) run
  automatic: ENTER_VALUE_HERE 
 
  retrieval:
    - # Which ranking approach do you use? E.g., bm25
      name: ENTER_VALUE_HERE

      # How many documents/chunks did you used in generation? Please put a number here, e.g., 10 or 1000
      documents: ENTER_VALUE_HERE

      ##################################################
      # Yes/No Questions
      ##################################################

      # Did you use any document chunking? (please enter: true/false)
      chunking: ENTER_VALUE_HERE 
  
      # Did you retrieve/re-rank chunks? (please enter: true/false)
      chunk_ranking: ENTER_VALUE_HERE

      # Did you use more than a single retrieval/ranking model? (please enter: true/false)
      single_stage_retrieval: ENTER_VALUE_HERE
      
  generation:
    # Please provide a short description of your generation approach.
    description: |
      ENTER_VALUE_HERE
      ...
      This is a multi line string
      ...
      ENTER_VALUE_HERE

    # Please provide a prompt use to generate answer.
    prompt: |
      ENTER_VALUE_HERE
      ...
      This is a multi line string
      ...
      ENTER_VALUE_HERE

    # Which LLM did you use in generation?
    llm: ENTER_VALUE_HERE 

    ##################################################
    # Yes/No Questions
    ##################################################

    # Did you use any reasoning? (please enter: true/false)
    reasoning: ENTER_VALUE_HERE 

    # Did you use any agentic approach? (please enter: true/false)
    agents: ENTER_VALUE_HERE