Shared Task
AINL:
ARTIFICIAL INTELLIGENCE AND NATURAL LANGUAGE CONFERENCE
ARTIFICIAL INTELLIGENCE AND NATURAL LANGUAGE CONFERENCE
•
•
•
INtroduction
Recent advancements in large language models (LLMs) have made them increasingly accessible across various domains.
Despite this progress, generating outputs from these models remains computationally intensive and costly, particularly due to lengthy input prompts.
Given that certain parts of an input prompt might contribute minimally to the output's relevance, optimizing the prompt structure becomes crucial for efficiency without compromising response quality.
Useful Links
Telegram chat: t.me/ainleval2026
Competition Platform: AINL-Eval 2026
Repo with baselines: github.com/iis-research-team/AINL-Eval-2026
Despite this progress, generating outputs from these models remains computationally intensive and costly, particularly due to lengthy input prompts.
Given that certain parts of an input prompt might contribute minimally to the output's relevance, optimizing the prompt structure becomes crucial for efficiency without compromising response quality.
Useful Links
Telegram chat: t.me/ainleval2026
Competition Platform: AINL-Eval 2026
Repo with baselines: github.com/iis-research-team/AINL-Eval-2026
Application form is under construction. Please return in a week or two.
Task Description
Participants must devise methods capable of compressing input prompts while preserving or even improving the quality of model-generated outputs.
Specifically, the challenge involves reducing the size of provided prompts while ensuring accurate generation of high-quality responses.
Submission
Please, refer to our GitHub to find a guide for submission process.
Evaluation
The target metrics:
If the overall length of submission (in chars) is not less than the initial lengths, then the answers evaluation won't be run and the Score will be 0.0 by default.
The team achieving the highest Overall Score wins.
Participation Rules
Final results will be obtained on the private test set and announced during the AINL 2026 conference.
Each team will have 3 attempts to submit final results.
After the competition is over, we will publish the ground truths for the test set, so the participants could perform some ablation studies.
Specifically, the challenge involves reducing the size of provided prompts while ensuring accurate generation of high-quality responses.
Submission
Please, refer to our GitHub to find a guide for submission process.
Evaluation
The target metrics:
- Compression Ratio: length of compressed prompts divided by original prompt length (characters).
- Score: proportion of correct answers produced when running the model with your optimized prompts.
- Overall Score: Compression Ratio x Score.
If the overall length of submission (in chars) is not less than the initial lengths, then the answers evaluation won't be run and the Score will be 0.0 by default.
The team achieving the highest Overall Score wins.
Participation Rules
Final results will be obtained on the private test set and announced during the AINL 2026 conference.
Each team will have 3 attempts to submit final results.
After the competition is over, we will publish the ground truths for the test set, so the participants could perform some ablation studies.
important dates
5 March 2026 - Competion starts
27 March 2026 - Test phase is open
3 April 2026 – The shared task is closed.
17-18 April 2026 – AINL 2026 conference. The final results are announced.
27 March 2026 - Test phase is open
3 April 2026 – The shared task is closed.
17-18 April 2026 – AINL 2026 conference. The final results are announced.
Competition Starts
March 5
March 27
Test Phase
is Open
is Open
Result
Announcement
Announcement
April 18
Test Phase
is Closed
is Closed
April 3
May 5
Paper Submission
Shared Task Organizers
Elena Bruches, Ershov Institute of Informatics Systems SB RAS
and Novosibirsk State University
Valentin Malykh, Moscow Institute of Physics and Technology
Tatiana Batura, Ershov Institute of Informatics Systems SB RAS
and Novosibirsk State University
Valentin Malykh, Moscow Institute of Physics and Technology
Tatiana Batura, Ershov Institute of Informatics Systems SB RAS
organizers
Contact us at ainlevent@gmail.com