Submission and Evaluation

You can review the results through our evaluation portal (https://poleval.amueval.pl/challenges), but here are the medallists:

Task 1a: ŚMIGIEL (Unsupervised)

  1. damian96 (0.7977)
  2. kwrobel (0.7574)

Task 1b: ŚMIGIEL (Constrained)

  1. kondziu98 (0.9253)
  2. grzmot (0.9127)
  3. tomek (0.9103)

Task 1c: ŚMIGIEL (Open)

  1. grzmot (0.8551)

Task 2a: Gender-inclusive LLMs for Polish (Proofreading)

  1. adam.majczyk (60.3933)
  2. chyul (9.9589)
  3. kwrobel (7.9424)

Task 2b: Gender-inclusive LLMs for Polish (Translation)

  1. kwrobel (69.3687)
  2. adam.majczyk (30.2884)
  3. mczajka (20.7871)

Task 4: Polish Speech Emotion Recognition Challenge

  1. maciejlachut (0.5412)
  2. tomasz (0.5247)
  3. tomek (0.5129)

FAQ

  1. Do I have to release the source code of my system to participate in PolEval?
    No, but you are strongly encouraged to do so. Releasing the source code is a strict condition for receiving the award.
  2. How many sets of answers can I submit?
    One participant (individual or team) may submit any number of systems, but up to 3 different sets of answers for each task or subtask will be evaluated (last 3 submitted). Those answers may come from different versions of your systems, different systems or models, systems with differently fine-tuned parameters, etc.