Discord

Checklist VsOSh AI 2026 Final Stage, Tour 1 (theory) · F task

A Fictional Situation

Russian title: Вымышленная ситуация

Decide whether GPU clusters can be allocated to 1,000 finalists without envy, given admissible supply triples and participants' linear preferences.

  • Proof (fair allocation / envy-freeness)
  • Russian original · English translation

The task

Each of 1,000 finalists receives a personal cluster of 8 devices of three types (green CUDA devices, red open-standard devices, yellow TPUs). A supply triple (y₁, y₂, y₃) is admissible if y₁ + y₂ + y₃ = 8000, every y_t ≥ 2026 and y₁ is a multiple of 7 (type 1 is delivered in batches of 7). Participant i values a cluster S = (x₁, x₂, x₃) at V_i(S) = u_{i,1}x₁ + u_{i,2}x₂ + u_{i,3}x₃, with preferences u given in gpu.csv (1,000 rows). An allocation is fair if V_i(S_i) ≥ V_i(S_j) for all participants i, j.

(a) Give an admissible triple for which no fair allocation exists, and prove it. (b) Does an admissible triple exist for which a fair allocation exists?

Abridged and translated by SOTA from the official Russian materials. The official statement has the exact rules, and it wins wherever this summary differs.

In English

This task was published in Russian. SOTA translated its 4 files into English on 17 September 2026.

Read the task statement in English 576 words

A Fictional Situation

English translation by SOTA – AI Community of the Russian original. Organisers who would like this translation removed can email [email protected].

All-Russian School Olympiad in Informatics 2025–2026, Final Stage
“Artificial Intelligence” profile, Tour 1, Moscow, 23 March 2026
Task F

Artur and Tanya are preparing computing resources for the practical tour for the 1000 participants of the final of the All-Russian School Olympiad in AI. Initially, they wanted to deploy the whole infrastructure in the cloud and hand out graphics cards (GPUs) from a common pool, but, to rule out network latency, it was decided to assemble a personal cluster of 8 devices for each participant.

The warehouse of the olympiad's partners holds 8000 devices of each of three types.

  • Type 1 (green): CUDA, fast but hot.
  • Type 2 (red): support open standards and are suitable for training large language models (LLMs), but require complicated driver set-up.
  • Type 3 (yellow): experimental accelerators (TPUs).

Under the terms of delivery, devices of type 1 can be obtained only in batches of 7; there must be at least 2026 devices of each type; exactly 8000 devices in total must be taken from the warehouse. If, in accordance with these conditions, y1y_1 devices of the first type, y2y_2 of the second and y3y_3 of the third can be brought from the warehouse, we call the triple of integers (y1,y2,y3)(y_1, y_2, y_3) admissible, that is:

y1+y2+y3=8000,yt2026 (t=1,2,3),y17.y_1 + y_2 + y_3 = 8000, \qquad y_t \geq 2026 \ (t = 1, 2, 3), \qquad y_1 \,\vdots\, 7.

(Translator's note: y17y_1 \,\vdots\, 7 is the Russian notation for "y1y_1 is divisible by 7".)

Each participant has a personal code i=1,2,,1000i = 1, 2, \ldots, 1000. The devices delivered from the warehouse are distributed among the participants. Let participant ii receive a personal cluster Si=(xi,1,xi,2,xi,3)S_i = (x_{i,1}, x_{i,2}, x_{i,3}), in which xi,1,xi,2,xi,3x_{i,1}, x_{i,2}, x_{i,3} are the numbers of devices of the first, second and third type respectively. The numbers xi,1,xi,2,xi,3x_{i,1}, x_{i,2}, x_{i,3} are non-negative integers, and xi,1+xi,2+xi,3=8x_{i,1} + x_{i,2} + x_{i,3} = 8.

It turned out that some participants prefer the "greens", some are fans of open standards and choose the "reds", and some love the unusual "yellows". For each participant, device preferences ui,1,ui,2,ui,3u_{i,1}, u_{i,2}, u_{i,3} are given; they are listed in the input file gpu.csv.

The file contains a header and 1000 rows with 3 columns of data: the ii-th row contains three integers ui,1,ui,2,ui,3u_{i,1}, u_{i,2}, u_{i,3} (in exactly this order), corresponding to the preferences of participant ii.

The utility of a cluster S=(x1,x2,x3)S = (x_1, x_2, x_3) for participant ii is given by the formula

Vi(S)=ui,1x1+ui,2x2+ui,3x3.V_i(S) = u_{i,1}x_1 + u_{i,2}x_2 + u_{i,3}x_3.

A distribution of devices is called fair if the inequality Vi(Si)Vi(Sj)V_i(S_i) \geq V_i(S_j) holds for any two participants i,ji, j.

(a) Give an example of an admissible triple of values (y1,y2,y3)(y_1, y_2, y_3) for which no fair distribution exists, and prove this.

(b) Does there exist an admissible triple (y1,y2,y3)(y_1, y_2, y_3) for which a fair distribution exists?

Translated by SOTA. The Russian original is the official version and wins wherever the two differ. The statement refers to the input file gpu.csv, which is not included in the PDF. The official solution contains two typos kept from the original (2087 for 2081, and b_1 − b'_2 for b_2 − b'_2 in Lemma 2). If you organise this olympiad and would like the translation removed, email [email protected] and we will take it down.

At a glance

You get
The file gpu.csv (header and 1,000 rows of three integers u_{i,1}, u_{i,2}, u_{i,3}).
You submit
Written solution (proof or answer with justification), graded by the criteria in the official solutions file.
Scoring
Points awarded by the published criteria (partial credit for stated intermediate results; deductions for gaps).
Rules
  • Written tour; no computer-based submission.
Format
Final stage, Tour 1 (theoretical), Moscow, 23 March 2026; grades 9–11; individual written tour.

Details

Year
2026, Moscow, Russia
Round
Final Stage, Tour 1 (theory) · F task
Language
Russian; English translation by SOTA
License
Not stated by the source