Free CAT library

Grouping & Distribution: Concepts & Solved Examples

Master the architectural principles of Grouping, Selection, Multi-Attribute Matching Matrices, and Capacity-Constrained Partitioning in CAT DILR. Formulate propositional selection logic (mutually exclusive pairs, concomitant inclusions, implication gates), construct high-dimensional cross-tabulation grids, deploy integer capacity bounds, and execute systematic constraint propagation pipelines.

1. Foundational Taxonomy of Grouping & Distribution Problems

In Data Interpretation and Logical Reasoning (DILR), Grouping & Distribution problems require partitioning, assigning, or selecting a finite set of entities E={e1,e2,…,en}\mathcal{E} = \{e_1, e_2, \dots, e_n\} into distinct groups, categories, or multi-attribute profiles subject to logical constraints, capacity bounds, and mutual dependencies.

Unlike pure Arrangement problems (which focus on linear, circular, or geometric slot coordinates), Grouping & Distribution puzzles center on set membership, bijective/injective matching, and combinatorial selection.

CAT-level problems in this domain broadly belong to three foundational archetypes:

  1. Combinatorial Team Selection: Selecting a subset of kk individuals from a larger pool of nn candidates (k<nk < n) under a web of conditional inclusion, mutual exclusion, and gender/domain representation quotas.

  2. Multi-Attribute Matching Matrices: Constructing a high-dimensional cross-tabulation grid to establish a one-to-one (bijective) correspondence across entities and multiple attribute tiers (e.g., Researcher ×\times Frontier Specialization ×\times Project Initiative ×\times Innovation Hub).

  3. Capacity-Constrained Partitioning: Distributing NN items or personnel across mm disjoint teams, buckets, or cohorts, where each group is bound by lower and upper integer capacity bounds (Lj≤∣Gj∣≤UjL_j \le |G_j| \le U_j), parity constraints, or summation quotas.


2. Mathematical Formalism of Selection Logic & Propositional Gates

In committee formation and team selection puzzles, clues are expressed in natural language that translates directly into propositional Boolean logic and algebraic 0-1 integer inequalities. Let xi∈{0,1}x_i \in \{0, 1\} denote the binary selection state of entity ii, where xi=1x_i = 1 indicates inclusion and xi=0x_i = 0 indicates exclusion.

Logical Grouping & Selection Dynamics

A. Mutually Exclusive Pair (Incompatibility)

  • Natural Language Statement: "Researcher A and Researcher B cannot both be selected for the task force."

  • Boolean Formulation: ¬(A∧B)  ⟺  ¬A∨¬B\neg (A \land B) \iff \neg A \lor \neg B

  • Algebraic 0-1 Inequality:

    xA+xB≤1x_A + x_B \le 1

  • Permissible Outcomes:

    • Select A only (xA=1,xB=0)(x_A = 1, x_B = 0)

    • Select B only (xA=0,xB=1)(x_A = 0, x_B = 1)

    • Select neither (xA=0,xB=0)(x_A = 0, x_B = 0)

  • Critical Invariant: Selecting A forces the exclusion of B (xA=1  ⟹  xB=0x_A = 1 \implies x_B = 0). However, rejecting A does NOT require selecting B.

B. Concomitant / Tied Selection (Inseparable Package Deal)

  • Natural Language Statement: "Researcher A and Researcher B must either both be selected, or neither can be selected."

  • Boolean Formulation: A  ⟺  BA \iff B

  • Algebraic 0-1 Equation:

    xA−xB=0  ⟺  xA=xBx_A - x_B = 0 \iff x_A = x_B

  • Permissible Outcomes: Both in the squad {A,B}\{A, B\}, or both excluded ∅\emptyset.

C. Unidirectional Implication Gate (Conditional Inclusion)

  • Natural Language Statement: "If Researcher A is selected, then Researcher B must also be selected."

  • Boolean Formulation: A  ⟹  B  ⟺  ¬A∨BA \implies B \iff \neg A \lor B

  • Algebraic 0-1 Inequality:

    xA≤xB  ⟺  xA−xB≤0x_A \le x_B \iff x_A - x_B \le 0

  • The Contrapositive Rule: ¬B  ⟹  ¬A\neg B \implies \neg A. If Researcher B is excluded from the squad, Researcher A CANNOT be selected.

  • The Dual Fallacy Warning:

    • Affirming the Consequent: Knowing B is selected (xB=1x_B = 1) does NOT imply A must be selected. A may or may not be chosen.

    • Denying the Antecedent: Knowing A is excluded (xA=0x_A = 0) does NOT imply B must be excluded. B can still be freely chosen.

D. Exclusive OR (Mandatory Alternation)

  • Natural Language Statement: "Exactly one of Researcher A or Researcher B must be selected."

  • Boolean Formulation: A⊕B  ⟺  (A∨B)∧¬(A∧B)A \oplus B \iff (A \lor B) \land \neg(A \land B)

  • Algebraic 0-1 Equation:

    xA+xB=1x_A + x_B = 1

  • Permissible Outcomes: Exactly one of the two candidates is chosen; both cannot be excluded, and both cannot be included.

Comprehensive Selection Rules Summary Matrix

Rule / Verbal ConditionBoolean LogicInteger FormulationValid States $(x_A, x_B)$Lethal Fallacy to Avoid
Mutual Exclusion ("Cannot both be selected")$\neg (A \land B)$$x_A + x_B \le 1$$(1, 0), (0, 1), (0, 0)$Assuming at least one must be picked
Tied Pair ("Both or neither")$A \iff B$$x_A = x_B$$(1, 1), (0, 0)$Selecting one while dropping the other
Conditional Inclusion ("If A then B")$A \implies B$$x_A \le x_B$$(1, 1), (0, 1), (0, 0)$Assuming $B \implies A$ or $\neg A \implies \neg B$
Either-Or Exclusive ("Exactly one")$A \oplus B$$x_A + x_B = 1$$(1, 0), (0, 1)$Allowing both to be dropped
Conditional Exclusion ("If A then not B")$A \implies \neg B$$x_A + x_B \le 1$$(1, 0), (0, 1), (0, 0)$Logically identical to Mutual Exclusion

3. Multi-Attribute Matching Matrices (Cross-Tabulation & Direct Elimination)

When a puzzle requires matching nn entities across kk categorical dimensions (e.g., nn Partners, nn Specializations, nn Innovation Hubs, nn Portfolio Startups), standard list representations fail due to combinatorial explosion. Instead, deploy the Cross-Tabulation Matrix Architecture.

The Master Entity Table vs. Pairwise Grid

  1. Primary Master Entity Table:

    Fix the principal entity list as rows (e.g., the nn individuals). Allocate separate columns for each secondary attribute.

Lead ArchitectFrontier SpecializationGlobal Innovation HubAssigned Compute Cluster
Architect AlphaMechanistic InterpretabilityZurichCluster-3
Architect BetaReinforcement LearningTokyoCluster-1
Architect GammaDiffusion PlanningSan FranciscoCluster-4
Architect DeltaNeuro-Symbolic ReasoningLondonCluster-2
  1. The Bijective Uniqueness Invariant:

    In any n×nn \times n sub-matrix mapping Attribute Domain X\mathcal{X} to Attribute Domain Y\mathcal{Y}:

    ∑j=1nMi,j=1(∀i),∑i=1nMi,j=1(∀j)\sum_{j=1}^n M_{i, j} = 1 \quad (\forall i), \qquad \sum_{i=1}^n M_{i, j} = 1 \quad (\forall j)

    Every row and every column must contain exactly one confirmed match (✓\checkmark). As soon as a cell (i,j)(i, j) is confirmed:

    • All other cells in row ii must be marked ×\times.

    • All other cells in column jj must be marked ×\times.

  2. The Principle of Transitive Linkage:

    If entity E1E_1 maps to attribute A2A_2 (E1→A2E_1 \to A_2), and attribute A2A_2 maps to attribute B3B_3 (A2→B3A_2 \to B_3), then by transitivity:

    E1→B3E_1 \to B_3

    Conversely, any negative constraint on B3B_3 (e.g., B3↛C4B_3 \not\to C_4) immediately propagates back to E1E_1 (E1↛C4E_1 \not\to C_4) and A2A_2 (A2↛C4A_2 \not\to C_4).


4. Capacity-Constrained Partitioning & Integer Extremal Bounds

In distribution puzzles, a total supply of NN discrete entities must be allocated across kk non-empty groups G1,G2,…,GkG_1, G_2, \dots, G_k such that:

∑j=1k∣Gj∣=N,Lj≤∣Gj∣≤Uj(∀j∈{1,2,…,k})\sum_{j=1}^k |G_j| = N, \qquad L_j \le |G_j| \le U_j \quad (\forall j \in \{1, 2, \dots, k\})

A. Min-Max Extremal Bound Theorem

To compute the extreme sizes of any target group GiG_i:

  • Theoretical Minimum: Push all other (k−1)(k - 1) groups to their upper capacity bounds UjU_j:

    ∣Gi∣min⁡=max⁡(Li,  N−∑j≠iUj)|G_i|_{\min} = \max\left(L_i, \; N - \sum_{j \ne i} U_j\right)

  • Theoretical Maximum: Push all other (k−1)(k - 1) groups to their lower capacity bounds LjL_j:

    ∣Gi∣max⁡=min⁡(Ui,  N−∑j≠iLj)|G_i|_{\max} = \min\left(U_i, \; N - \sum_{j \ne i} L_j\right)

B. Strictly Distinct Partition Envelopes

If a problem dictates that all kk group sizes must be strictly distinct integers (∣G1∣<∣G2∣<⋯<∣Gk∣|G_1| < |G_2| < \dots < |G_k|) with a minimum base size LL:

  • The minimum total elements required to sustain kk distinct non-empty groups is:

    Nmin⁡=∑j=0k−1(L+j)=k⋅L+k(k−1)2N_{\min} = \sum_{j=0}^{k-1} (L + j) = k \cdot L + \frac{k(k - 1)}{2}

    Example: For k=4k = 4 groups with minimum size L=3L = 3, the minimal sum is 3+4+5+6=183 + 4 + 5 + 6 = 18. If the actual total is N=21N = 21, the 3 surplus units must be distributed without violating the distinctness order.

Concrete Partition Bounding Example

Scenario ParameterMathematical ConditionDerived Integer Range
Total Entities ($N$)$\sum_{j=1}^4 |G_j| = 21$Fixed Sum $= 21$
Base Bounds$3 \le |G_j| \le 8$$\forall j \in \{1, 2, 3, 4\}$
Strict Hierarchy$|G_1| > |G_2| \ge 3$$|G_1| \ge 4$
Maximum of Group 1$|G_1|_{\max} = 21 - (3 + 3 + 3) = 12 \to \min(8, 12)$$|G_1|_{\max} = 8$
Maximum of Group 2$|G_2| \le 5$ (since $|G_1| \ge 6$ and remaining $\ge 10$)$|G_2|_{\max} = 5$

5. The Master 4-Phase Execution Pipeline


+-------------------------------------------------------------------+

| Phase 1: Problem Domain Modeling & Degree-of-Freedom Mapping      |
| - Identify problem archetype: Selection, Matrix Matching, Bucket.  |
| - Tally entities (N), group counts (k), and integer boundaries.   |

+-------------------------------------------------------------------+

                                  |

                                  v

+-------------------------------------------------------------------+

| Phase 2: Direct Anchor Placement & Definite Matchings             |
| - Insert unconditional clues directly into the Master Table.      |
| - Cross out row/column counterparts by Bijective Uniqueness.      |

+-------------------------------------------------------------------+

                                  |

                                  v

+-------------------------------------------------------------------+

| Phase 3: Constraint Compilation & Relational Filtering           |
| - Translate verbal clues into Boolean gates and inequalities.     |
| - Propagate transitive relationships across attribute tiers.      |

+-------------------------------------------------------------------+

                                  |

                                  v

+-------------------------------------------------------------------+

| Phase 4: Tree-Search Branching & Invariant Deduction              |
| - If multiple candidate states survive, identify the binary fork. |
| - Test Branch Alpha vs. Branch Beta; eliminate contradictions.    |
| - Verify all survivors against global parity and quotas.          |

+-------------------------------------------------------------------+

Formulas

Propositional Selection Inequality

xA≤xB  ⟺  A  ⟹  B,xA+xB≤1  ⟺  ¬(A∧B),xA=xB  ⟺  A  ⟺  Bx_A \le x_B \iff A \implies B, \quad x_A + x_B \le 1 \iff \neg(A \land B), \quad x_A = x_B \iff A \iff B

Translates verbal conditional selection statements into linear 0-1 integer inequalities. Implication imposes an upper bound, mutual exclusion caps the sum at 1, and tied pairs enforce equality.
Capacity-Constrained Extremal Bounds

∣Gi∣min⁡=max⁡(Li,  N−∑j≠iUj),∣Gi∣max⁡=min⁡(Ui,  N−∑j≠iLj)|G_i|_{\min} = \max\left(L_i, \; N - \sum_{j \ne i} U_j\right), \quad |G_i|_{\max} = \min\left(U_i, \; N - \sum_{j \ne i} L_j\right)

Computes the exact lower and upper integer limits for group G_i when distributing N entities across k groups with individual capacity constraints [L_j, U_j].
Multi-Attribute Bijective Invariant

∑i=1nMi,j=1,∑j=1nMi,j=1,(Ai↦Bj∧Bj↦Ck)  ⟹  Ai↦Ck\sum_{i=1}^n M_{i, j} = 1, \quad \sum_{j=1}^n M_{i, j} = 1, \quad (A_i \mapsto B_j \land B_j \mapsto C_k) \implies A_i \mapsto C_k

Enforces one-to-one matching across categorical domains in cross-tabulation grids. Confirmation of any cell immediately eliminates its row and column; linkages propagate transitively.
Strictly Distinct Partition Minimal Envelope

N≥∑j=0k−1(L+j)=k⋅L+k(k−1)2N \ge \sum_{j=0}^{k-1} (L + j) = k \cdot L + \frac{k(k - 1)}{2}

Determines the absolute minimum total elements required to construct k strictly distinct integer groups with minimum baseline group size L.

Traps and tips

  • Trap 1: The Implication Direction Fallacy (Affirming the Consequent) — When a clue states 'If Dr. Aris is selected, then Dr. Bianca must be selected' (A  ⟹  BA \implies B), students frequently commit the fatal error of assuming that selecting Dr. Bianca forces the selection of Dr. Aris. In reality, Bianca can be chosen while Aris is excluded. The ONLY valid logical deduction is the contrapositive: if Bianca is rejected, Aris cannot be selected ($

    eg B \implies \neg A$).

  • Trap 2: Confusing 'Mutually Exclusive' with 'Either-Or' — 'A and B cannot both be chosen' (xA+xB≤1x_A + x_B \le 1) permits THREE valid outcomes: A only, B only, or NEITHER. In contrast, 'Exactly one of A or B must be chosen' (xA+xB=1x_A + x_B = 1) allows only TWO outcomes. Never force the selection of an entity from a mutually exclusive pair unless explicitly required by squad size quotas.
  • Trap 3: Transitive Elimination Leakage in Multi-Attribute Grids — If a clue states 'The researcher from Tokyo specializes in Computer Vision' and a later clue states 'The Computer Vision specialist does not use Cluster-1', students often fail to link Tokyo with Cluster-1. By transitivity, the Tokyo researcher CANNOT use Cluster-1. Always propagate negative attributes across all linked dimensions.
  • Trap 4: Off-by-One Interval Errors in Capacity-Constrained Buckets — When clues state that 'Division X has more interns than Division Y' where Division Y has at least 3 interns, Division X must have at least 3+1=43 + 1 = 4 interns. Forgetting the strict inequality offset leads to illegal minimal allocations.
  • Tip 5: Prioritize Extreme Partition Anchors First — In distribution sets with bounded groups, calculate the extremal limits of the largest and smallest groups before evaluating secondary clues. Locking in ∣G1∣max⁡|G_1|_{\max} or ∣Gk∣min⁡|G_k|_{\min} immediately collapses the permissible integer partition tree.

Solved examples

Solved Example Set 1: Questions 1 to 4 (Frontier DeepTech AI Compute Foundations & Capital Allocation)

Directions for Questions 1 to 4

Four frontier DeepTech AI compute foundations—AetherAI, TensorCore, BioNeuro, and LuminaLLM—operate high-performance sovereign compute clusters in Silicon Valley. During the fiscal year FY 2025–26, their operational parameters across four distinct metrics were recorded:

  • Compute Base (in PFLOPS)

  • Enterprise Deployments (in $ Million)

  • Global Cloud API Exports (in $ Million)

  • Operating Net Surplus (in $ Million)

The foundation audit committee established the following systemic relationships:

  • Clue I: Enterprise deployments of LuminaLLM were exactly 12 times the operating net surplus of TensorCore.

  • Clue II: The sum of global cloud API exports of BioNeuro and LuminaLLM is equal to half the compute base of AetherAI.

  • Clue III: The sum of the compute base of BioNeuro and the enterprise deployments of TensorCore is equal to five times the global cloud API exports of TensorCore.

  • Clue IV: Operating net surplus of LuminaLLM = Enterprise deployments of AetherAI − Global cloud API exports of BioNeuro.

  • Clue V: The operating net surplus of AetherAI is equal to the arithmetic average of the operating net surpluses of TensorCore and LuminaLLM.

  • Clue VI: The total operating net surplus of all four foundations put together was $140 million.

  • Clue VII: Enterprise deployments of BioNeuro was 1.25 times the global cloud API exports of AetherAI.

  • Clue VIII: The compute base of TensorCore = Total global cloud API exports of all four foundations combined − 80 PFLOPS.

  • Clue IX: The compute base of LuminaLLM = 6×6 \times (Operating net surplus of LuminaLLM + Operating net surplus of BioNeuro − Operating net surplus of TensorCore).

Furthermore, the audit team confirmed the following specific baseline metrics:

  1. Enterprise deployments of LuminaLLM = $180 million

  2. Compute base of AetherAI = 600 PFLOPS

  3. Enterprise deployments of BioNeuro = $150 million

  4. Compute base of BioNeuro = 620 PFLOPS

  5. Global cloud API exports of TensorCore = $160 million

  6. Operating net surplus of BioNeuro = $50 million

  7. Global cloud API exports of BioNeuro = $120 million


Question 1

In FY 2025–26, what was the compute base of LuminaLLM?

(1) 450 PFLOPS

(2) 480 PFLOPS

(3) 500 PFLOPS

(4) 520 PFLOPS

Question 2

In FY 2025–26, what was the operating net surplus of LuminaLLM?

(1) $30 million

(2) $35 million

(3) $45 million

(4) $50 million

Question 3

In FY 2025–26, by what percentage did the compute base of AetherAI exceed the compute base of TensorCore?

(1) 15%

(2) 20%

(3) 25%

(4) 30%

Question 4

In FY 2025–26, the total enterprise deployments of all four foundations combined was approximately what percentage of the total compute base of all four foundations combined?

(1) 25%

(2) 31%

(3) 36%

(4) 42%

Show solution

Step-by-Step Deductive Solution

1. Formulating the Operational Matrix

Let the four foundations be denoted as AA (AetherAI), TT (TensorCore), BB (BioNeuro), and LL (LuminaLLM).

Let the four operational parameters be:

  • CC: Compute Base (PFLOPS)

  • DD: Enterprise Deployments ($M)

  • EE: Global Cloud API Exports ($M)

  • SS: Operating Net Surplus ($M)

From the baseline data given directly:

  • D(L)=180D(L) = 180

  • C(A)=600C(A) = 600

  • D(B)=150D(B) = 150

  • C(B)=620C(B) = 620

  • E(T)=160E(T) = 160

  • S(B)=50S(B) = 50

  • E(B)=120E(B) = 120


2. Sequential Propagation of Clues

  1. Surplus of TensorCore (Clue I):

    D(L)=12×S(T)  ⟹  180=12×S(T)  ⟹  S(T)=18012=$15 millionD(L) = 12 \times S(T) \implies 180 = 12 \times S(T) \implies S(T) = \frac{180}{12} = \mathbf{\text{\textdollar}15\text{ million}}

  2. Exports of LuminaLLM (Clue II):

    E(B)+E(L)=12C(A)  ⟹  120+E(L)=6002=300  ⟹  E(L)=300−120=$180 millionE(B) + E(L) = \frac{1}{2} C(A) \implies 120 + E(L) = \frac{600}{2} = 300 \implies E(L) = 300 - 120 = \mathbf{\text{\textdollar}180\text{ million}}

  3. Enterprise Deployments of TensorCore (Clue III):

    C(B)+D(T)=5×E(T)  ⟹  620+D(T)=5×160=800  ⟹  D(T)=800−620=$180 millionC(B) + D(T) = 5 \times E(T) \implies 620 + D(T) = 5 \times 160 = 800 \implies D(T) = 800 - 620 = \mathbf{\text{\textdollar}180\text{ million}}

  4. Exports of AetherAI (Clue VII):

    D(B)=1.25×E(A)  ⟹  150=1.25×E(A)  ⟹  E(A)=1501.25=$120 millionD(B) = 1.25 \times E(A) \implies 150 = 1.25 \times E(A) \implies E(A) = \frac{150}{1.25} = \mathbf{\text{\textdollar}120\text{ million}}

  5. Resolving Operating Surpluses (Clues V & VI):

    • Total surplus of all four foundations =140= 140 (Clue VI):

      S(A)+S(T)+S(B)+S(L)=140S(A) + S(T) + S(B) + S(L) = 140

      S(A)+15+50+S(L)=140  ⟹  S(A)+S(L)=75— (Eq. 1)S(A) + 15 + 50 + S(L) = 140 \implies S(A) + S(L) = 75 \quad \text{--- (Eq. 1)}

    • Surplus of AetherAI is the average of TensorCore and LuminaLLM (Clue V):

      S(A)=S(T)+S(L)2=15+S(L)2  ⟹  2S(A)−S(L)=15— (Eq. 2)S(A) = \frac{S(T) + S(L)}{2} = \frac{15 + S(L)}{2} \implies 2S(A) - S(L) = 15 \quad \text{--- (Eq. 2)}

    • Adding (Eq. 1) and (Eq. 2):

      3S(A)=75+15=90  ⟹  S(A)=$30 million3S(A) = 75 + 15 = 90 \implies S(A) = \mathbf{\text{\textdollar}30\text{ million}}

      S(L)=75−30=$45 millionS(L) = 75 - 30 = \mathbf{\text{\textdollar}45\text{ million}}

  6. Deployments of AetherAI (Clue IV):

    S(L)=D(A)−E(B)  ⟹  45=D(A)−120  ⟹  D(A)=120+45=$165 millionS(L) = D(A) - E(B) \implies 45 = D(A) - 120 \implies D(A) = 120 + 45 = \mathbf{\text{\textdollar}165\text{ million}}

  7. Compute Base of TensorCore (Clue VIII):

    • Total exports of all four foundations:

      Etotal=E(A)+E(T)+E(B)+E(L)=120+160+120+180=580 $ME_{\text{total}} = E(A) + E(T) + E(B) + E(L) = 120 + 160 + 120 + 180 = 580\text{ \text{\textdollar}M}

    • Therefore:

      C(T)=Etotal−80=580−80=500 PFLOPSC(T) = E_{\text{total}} - 80 = 580 - 80 = \mathbf{500\text{ PFLOPS}}

  8. Compute Base of LuminaLLM (Clue IX):

    C(L)=6×(S(L)+S(B)−S(T))=6×(45+50−15)=6×80=480 PFLOPSC(L) = 6 \times (S(L) + S(B) - S(T)) = 6 \times (45 + 50 - 15) = 6 \times 80 = \mathbf{480\text{ PFLOPS}}


Master Foundation Performance Matrix

Operational MetricAetherAITensorCoreBioNeuroLuminaLLMTotal Portfolio
Compute Base (PFLOPS)6005006204802,200
Enterprise Deployments ($M)165180150180675
Global Cloud Exports ($M)120160120180580
Operating Net Surplus ($M)30155045140

Question-by-Question Answers

Question 1:

  • From the table, the compute base of LuminaLLM is C(L)=480 PFLOPSC(L) = \mathbf{480\text{ PFLOPS}}.

  • Correct Choice: (2) 480 PFLOPS

Question 2:

  • From the table, the operating net surplus of LuminaLLM is S(L)=$45 millionS(L) = \mathbf{\text{\textdollar}45\text{ million}}.

  • Correct Choice: (3) $45 million

Question 3:

  • Compute base of AetherAI =600= 600 PFLOPS.

  • Compute base of TensorCore =500= 500 PFLOPS.

  • Percentage excess:

    600−500500×100%=100500×100%=20%\frac{600 - 500}{500} \times 100\% = \frac{100}{500} \times 100\% = \mathbf{20\%}

  • Correct Choice: (2) 20%

Question 4:

  • Total enterprise deployments =165+180+150+180=$675 million= 165 + 180 + 150 + 180 = \text{\textdollar}675\text{ million}.

  • Total compute base =600+500+620+480=2,200 PFLOPS= 600 + 500 + 620 + 480 = 2,200\text{ PFLOPS}.

  • Required percentage:

    6752200×100%≈30.68%≈31%\frac{675}{2200} \times 100\% \approx 30.68\% \approx \mathbf{31\%}

  • Correct Choice: (2) 31%

Solved Example Set 2: Questions 5 to 8 (Advanced AI Alignment & Safety Roundtable Deliberation)

Directions for Questions 5 to 8

Six principal research fellows—Dr. Anya, Dr. Braxton, Dr. Clarissa, Dr. Declan, Dr. Esha, and Dr. Florian—participate in an executive AI Alignment Summit. They sit in a single block of two opposing parallel rows of three seats each (North Row facing South, and South Row facing North), such that each fellow directly faces a fellow in the opposite row.

Each fellow has authored a seminal monograph in her/his primary specialization domain, and each is currently reviewing a technical draft paper written by one of the other five fellows. No two fellows are reviewing a paper in the same specialization domain.

The six specialization domains represented are:

  • Mechanistic Interpretability

  • RLHF (Reinforcement Learning from Human Feedback)

  • Bio-Digital Neural Interfaces

  • Formal Verification & Safety Proofs

  • Cognitive Agent Psychology

  • Quantum Neural Optimization

The symposium chairs note the following seating and review protocol constraints:

  • Clue I: Dr. Anya, who occupies a corner seat of a row, is reviewing the draft paper on Mechanistic Interpretability. Her own authored monograph is not on Cognitive Agent Psychology.

  • Clue II: Dr. Braxton is reviewing the draft paper authored by the fellow seated directly opposite him/her.

  • Clue III: Dr. Clarissa, seated in the middle seat between the fellows whose authored monographs are on Mechanistic Interpretability and RLHF, is reviewing the draft paper on Bio-Digital Neural Interfaces.

  • Clue IV: Dr. Declan, seated directly opposite the fellow whose authored monograph is on Bio-Digital Neural Interfaces, is reviewing the draft paper on Formal Verification & Safety Proofs.

  • Clue V: The fellow whose authored monograph is on Mechanistic Interpretability sits directly opposite the fellow whose authored monograph is on Cognitive Agent Psychology.

  • Clue VI: Dr. Esha, seated adjacent to the fellow whose authored monograph is on Formal Verification & Safety Proofs, is reviewing the draft paper on RLHF.

  • Clue VII: Dr. Florian, who occupies a corner seat of a row, is seated adjacent to the fellow whose authored monograph is on Quantum Neural Optimization.


Question 5

Who is seated directly opposite Dr. Clarissa?

(1) Dr. Anya

(2) Dr. Braxton

(3) Dr. Esha

(4) Dr. Declan

Question 6

Which draft paper is Dr. Florian reviewing?

(1) Cognitive Agent Psychology

(2) RLHF

(3) Quantum Neural Optimization

(4) Bio-Digital Neural Interfaces

Question 7

What is Dr. Anya's authored specialization domain?

(1) Quantum Neural Optimization

(2) Formal Verification & Safety Proofs

(3) Bio-Digital Neural Interfaces

(4) Cognitive Agent Psychology

Question 8

Which of the following pairs of fellows are NOT seated directly opposite each other?

(1) Dr. Anya & Dr. Declan

(2) Dr. Braxton & Dr. Clarissa

(3) Dr. Esha & Dr. Florian

(4) None of these (all given pairs are seated directly opposite)

Show solution

Step-by-Step Deductive Solution

1. Table Topology & Degree of Freedom Mapping

Let the seating arrangement comprise two facing rows of three seats:

  • Row 1 (North Row): Seats 1, 2, 3 (facing South). Seat 2 is the middle seat; Seats 1 and 3 are corner seats.

  • Row 2 (South Row): Seats 4, 5, 6 (facing North). Seat 5 is the middle seat; Seats 4 and 6 are corner seats.

  • Opposing seat pairs: (1,4),(2,5),(3,6)(1, 4), (2, 5), (3, 6).

Each fellow has two categorical attributes:

  1. Authored Monograph Domain (own specialization)

  2. Draft Paper Being Reviewed (must differ from own domain, and all 6 must be distinct)


2. Deducing Middle Seats & Opposing Structure

  1. Locating Dr. Clarissa (Clue III):

    • Dr. Clarissa sits between two fellows   ⟹  \implies Clarissa MUST occupy a middle seat (Seat 2 or Seat 5).

    • Without loss of generality, place Dr. Clarissa at Seat 5 (middle of Row 2).

    • Clarissa is reviewing Bio-Digital Neural Interfaces.

    • Her adjacent neighbors in Row 2 (Seats 4 and 6) authored Mechanistic Interpretability and RLHF.

  2. Locating Dr. Florian & Clarissa's Authored Domain (Clue VII):

    • Dr. Florian sits at a corner seat adjacent to the fellow who authored Quantum Neural Optimization.

    • The middle seat of Row 2 is Clarissa; the middle seat of Row 1 is Seat 2.

    • If Florian is in Row 2 (Seat 4 or 6), his adjacent middle seat is Clarissa.

    • This implies Dr. Clarissa's authored monograph is Quantum Neural Optimization!

    • Florian occupies one of the corner seats of Row 2 (say Seat 6), meaning the fellow who authored Mechanistic Interpretability is at Seat 6, so Dr. Florian authored Mechanistic Interpretability!

    • Then Seat 4 must be the fellow who authored RLHF.

  3. Locating Dr. Braxton (Clue II):

    • Dr. Braxton reviews the paper authored by the fellow sitting directly opposite him/her.

    • Clarissa authored Quantum Neural Optimization. If Braxton sits directly opposite Clarissa (at Seat 2, the middle seat of Row 1), Braxton reviews Quantum Neural Optimization.

    • This satisfies Clue II perfectly: Dr. Braxton is at Seat 2!

  4. Locating Dr. Declan (Clue IV):

    • Dr. Declan reviews Formal Verification & Safety Proofs.

    • Declan sits directly opposite the fellow who authored Bio-Digital Neural Interfaces.

    • Neither Seat 2 (Braxton) nor Seat 5 (Clarissa) is Declan.

    • In Row 2, Seats 4 and 6 authored RLHF and Mechanistic Interpretability.

    • Thus, the fellow who authored Bio-Digital Neural Interfaces must sit in Row 1!

    • Declan must sit in Row 2 opposite this person.

    • Since Seat 6 is Florian, Dr. Declan must occupy Seat 4!

    • Since Declan is at Seat 4, his opposite seat (Seat 1 in Row 1) must be the fellow who authored Bio-Digital Neural Interfaces.

    • Also, since Seat 4 is Declan, Declan authored RLHF!

  5. Identifying Dr. Anya (Clue I & V):

    • The fellow who authored Mechanistic Interpretability (Florian, Seat 6) sits directly opposite the fellow who authored Cognitive Agent Psychology (Clue V).

    • The seat opposite Seat 6 is Seat 3 in Row 1. Thus, Seat 3 authored Cognitive Agent Psychology.

    • Remaining fellow for Seat 1 (corner of Row 1) is Dr. Anya!

    • Since Seat 1 authored Bio-Digital Neural Interfaces, Dr. Anya authored Bio-Digital Neural Interfaces!

    • Anya reviews Mechanistic Interpretability (Clue I), and she does not have interest in Cognitive Agent Psychology (verified!).

  6. Completing Row 1 with Dr. Esha (Clue VI):

    • Seat 3 must be occupied by Dr. Esha!

    • Esha authored Cognitive Agent Psychology and reviews RLHF (Clue VI).

    • Esha sits adjacent to Seat 2 (Braxton). Clue VI states Esha is seated adjacent to the fellow who authored Formal Verification & Safety Proofs   ⟹  \implies Dr. Braxton authored Formal Verification & Safety Proofs!

  7. Determining Florian's Review Paper:

    • The six review domains are: Mechanistic (Anya), Quantum (Braxton), Bio-Digital (Clarissa), Formal (Declan), RLHF (Esha).

    • The sole remaining review domain is Cognitive Agent Psychology.

    • Therefore, Dr. Florian reviews Cognitive Agent Psychology!


Master Roundtable Configuration Matrix

North Row (Facing South ↓)
Seat PositionResearch FellowAuthored Monograph DomainDraft Paper Being Reviewed
Seat 1 (Corner)Dr. AnyaBio-Digital Neural InterfacesMechanistic Interpretability
Seat 2 (Middle)Dr. BraxtonFormal Verification & Safety ProofsQuantum Neural Optimization
Seat 3 (Corner)Dr. EshaCognitive Agent PsychologyRLHF
↔ Direct Facing Line of Sight ↔
Seat 4 (Corner)Dr. DeclanRLHFFormal Verification & Safety Proofs
Seat 5 (Middle)Dr. ClarissaQuantum Neural OptimizationBio-Digital Neural Interfaces
Seat 6 (Corner)Dr. FlorianMechanistic InterpretabilityCognitive Agent Psychology
South Row (Facing North ↑)

Question-by-Question Answers

Question 5:

  • Dr. Clarissa occupies Seat 5 (middle of South Row); the seat directly opposite is Seat 2, occupied by Dr. Braxton.

  • Correct Choice: (2) Dr. Braxton

Question 6:

  • From the table, Dr. Florian is reviewing Cognitive Agent Psychology (authored by Dr. Esha, who sits directly opposite him).

  • Correct Choice: (1) Cognitive Agent Psychology

Question 7:

  • Dr. Anya authored the monograph on Bio-Digital Neural Interfaces.

  • Correct Choice: (3) Bio-Digital Neural Interfaces

Question 8:

  • Anya and Declan sit directly opposite each other (Seats 1 and 4).

  • Braxton and Clarissa sit directly opposite each other (Seats 2 and 5).

  • Esha and Florian sit directly opposite each other (Seats 3 and 6).

  • Therefore, all three given pairs are seated directly opposite each other.

  • Correct Choice: (4) None of these (all given pairs are seated directly opposite)

Solved Example Set 3: Questions 9 to 12 (Global DeepTech Venture Syndicate Leads & Portfolio Deployments)

Directions for Questions 9 to 12

Six venture capital partners—Garrick, Julian, Armand, Dylan, Soren, and Ronan—lead deeptech investments across six global innovation hubs: Zurich, Berlin, Tokyo, Singapore, Austin, and Bangalore.

Each partner is closely paired with a distinct portfolio CTO from among six founders: Amara, Kaia, Jacqueline, Nadia, Maya, and Dana. Furthermore, each partner focuses primarily on a distinct frontier technology domain: Quantum Computing, Autonomous Robotics, Generative AI, Synthetic Biology, Clean Fusion, and Cybersecurity.

The syndicate operations agreement specifies the following facts:

  • Clue I: The venture partners collaborating with portfolio CTOs Jacqueline, Maya, or Dana do not lead investments in Autonomous Robotics or Generative AI.

  • Clue II: The venture partner based in the Tokyo innovation hub leads investments in Quantum Computing.

  • Clue III: Ronan, who operates out of the Bangalore innovation hub, leads investments in Autonomous Robotics.

  • Clue IV: Armand and Soren, neither of whom is based in Tokyo, partner with portfolio CTOs Amara and Nadia respectively.

  • Clue V: The venture partners operating out of Berlin and Singapore lead investments in Clean Fusion and Generative AI respectively.

  • Clue VI: Dylan operates out of the Austin innovation hub.

  • Clue VII: Armand leads investments in Synthetic Biology.


Question 9

The venture partner operating out of the Singapore innovation hub is partnered with which portfolio CTO?

(1) Jacqueline

(2) Nadia

(3) Amara

(4) Cannot be determined

Question 10

The venture partner who leads investments in Synthetic Biology operates out of which innovation hub?

(1) Austin

(2) Zurich

(3) Bangalore

(4) Cannot be determined

Question 11

Which frontier technology domain is led by Dylan?

(1) Cybersecurity

(2) Autonomous Robotics

(3) Generative AI

(4) Cannot be determined

Question 12

If Maya's venture partner leads investments in Clean Fusion and Dana is partnered with Garrick, then Dylan's portfolio CTO partner is:

(1) Maya

(2) Kaia

(3) Jacqueline

(4) None of these

Show solution

Step-by-Step Deductive Solution

1. Attribute Domains & Fixed Linkages

The problem establishes a 4-dimensional bijective matching:

  • Partners (6): Garrick, Julian, Armand, Dylan, Soren, Ronan

  • Portfolio CTOs (6): Amara, Kaia, Jacqueline, Nadia, Maya, Dana

  • Innovation Hubs (6): Zurich, Berlin, Tokyo, Singapore, Austin, Bangalore

  • Frontier Domains (6): Quantum Computing, Autonomous Robotics, Generative AI, Synthetic Biology, Clean Fusion, Cybersecurity

From Clues II, III, and V, we establish direct hub-to-domain linkages:

  • Tokyo ↔\leftrightarrow Quantum Computing

  • Bangalore ↔\leftrightarrow Autonomous Robotics (led by Ronan, Clue III)

  • Berlin ↔\leftrightarrow Clean Fusion

  • Singapore ↔\leftrightarrow Generative AI


2. Deducing Partners, Domains, and Hubs

  1. Ronan's Full Profile (Clue III):

    • Partner: Ronan

    • Hub: Bangalore

    • Domain: Autonomous Robotics

  2. Armand's Profile (Clues IV & VII):

    • Partner: Armand

    • Portfolio CTO: Amara (Clue IV)

    • Domain: Synthetic Biology (Clue VII)

    • Hub: Armand is not Tokyo (Clue IV). Hubs with pre-assigned domains are Tokyo (Quantum), Bangalore (Robotics), Berlin (Fusion), Singapore (GenAI). Since Armand leads Synthetic Biology, he cannot be Berlin or Singapore.

    • Dylan is at Austin (Clue VI).

    • Thus, the ONLY remaining hub available for Armand is Zurich!

  3. Soren's Profile (Clues I, IV, & V):

    • Partner: Soren

    • Portfolio CTO: Nadia (Clue IV)

    • Clue I states that partners of Jacqueline, Maya, and Dana CANNOT lead Autonomous Robotics or Generative AI.

    • Therefore, the two partners leading Autonomous Robotics and Generative AI MUST be partnered with CTOs from the remaining set: {Amara,Kaia,Nadia}\{\text{Amara}, \text{Kaia}, \text{Nadia}\}.

    • We know Ronan leads Robotics (partnered with Kaia), and Armand is partnered with Amara (Synthetic Biology).

    • Consequently, the partner leading Generative AI MUST be Soren!

    • From Clue V, the Generative AI lead operates out of Singapore.

    • Therefore: Soren is based in Singapore, leads Generative AI, and is partnered with Nadia!

  4. Ronan's Portfolio CTO:

    • Ronan leads Autonomous Robotics and must be partnered with a founder from {Amara,Kaia,Nadia}\{\text{Amara}, \text{Kaia}, \text{Nadia}\}.

    • Since Armand is with Amara and Soren is with Nadia, Ronan MUST be partnered with Kaia!

  5. Dylan's Domain (Clue VI):

    • Partner: Dylan

    • Hub: Austin (Clue VI)

    • The domains linked to other hubs are: Quantum (Tokyo), Robotics (Bangalore), Fusion (Berlin), GenAI (Singapore), Synthetic Biology (Zurich).

    • The only remaining domain is Cybersecurity.

    • Therefore: Dylan leads Cybersecurity from Austin!

  6. Resolving Garrick and Julian:

    • The two remaining partners are Garrick and Julian.

    • The two remaining hubs are Tokyo (Quantum Computing) and Berlin (Clean Fusion).

    • Thus, Garrick and Julian divide {Tokyo/Quantum,Berlin/Fusion}\{\text{Tokyo/Quantum}, \text{Berlin/Fusion}\}.

    • The three remaining portfolio CTOs to be distributed among Dylan, Garrick, and Julian are {Jacqueline,Maya,Dana}\{\text{Jacqueline}, \text{Maya}, \text{Dana}\}.


Master Venture Syndicate Matrix

Venture PartnerPortfolio CTOInnovation HubFrontier Technology Domain
GarrickJacqueline / Maya / DanaBerlin / TokyoClean Fusion / Quantum Computing
JulianJacqueline / Maya / DanaTokyo / BerlinQuantum Computing / Clean Fusion
ArmandAmaraZurichSynthetic Biology
DylanJacqueline / Maya / DanaAustinCybersecurity
SorenNadiaSingaporeGenerative AI
RonanKaiaBangaloreAutonomous Robotics

Question-by-Question Answers

Question 9:

  • The partner operating out of Singapore is Soren, who is partnered with Nadia.

  • Correct Choice: (2) Nadia

Question 10:

  • Armand leads Synthetic Biology and operates out of Zurich.

  • Correct Choice: (2) Zurich

Question 11:

  • Dylan is based in Austin and leads Cybersecurity.

  • Correct Choice: (1) Cybersecurity

Question 12:

  • If Maya's partner leads Clean Fusion (Berlin), then Maya is paired with whichever partner is at Berlin.

  • If Dana is partnered with Garrick, then Garrick cannot be with Jacqueline.

  • Since Maya and Dana are assigned to the two partners dividing Berlin and Tokyo (Garrick and Julian), the sole remaining CTO for Dylan is Jacqueline!

  • Correct Choice: (3) Jacqueline

Solved Example Set 4: Questions 13 to 16 (Global LLM Agent Benchmark Evaluation Tiers)

Directions for Questions 13 to 16

Four autonomous AI reasoning agent models—Javelin, Kestrel, Lynx, and Maven—competed in the annual Global LLM Agent Benchmark (Agent Arena 2026), capturing the top four ranks (Rank 1 to Rank 4).

Each agent model was engineered by a distinct sponsoring deeptech lab: Apex Systems, DeepDynamix, Synapse Core, and Radiant Intelligence. Furthermore, each agent model is constructed upon a distinct state-of-the-art neural architecture:

  • Autoregressive Transformer

  • Diffusion Planner Engine

  • State-Space Mamba Hybrid

  • Neuro-Symbolic Reasoner

The benchmark arbitration panel released the following performance telemetry:

  • Clue I: The agent sponsored by Synapse Core finished neither first nor second.

  • Clue II: The agent built on the State-Space Mamba Hybrid architecture finished first (Rank 1).

  • Clue III: Agent Maven finished second (Rank 2).

  • Clue IV: Agent Javelin is built on the Autoregressive Transformer architecture.

  • Clue V: The agent sponsored by DeepDynamix, which is built on the Neuro-Symbolic Reasoner architecture, ranked fourth (Rank 4).

  • Clue VI: The name of the agent sponsored by Radiant Intelligence is Kestrel.


Question 13

The first rank (Rank 1) in the benchmark was captured by:

(1) The agent sponsored by Apex Systems

(2) The agent sponsored by Radiant Intelligence

(3) Maven

(4) Javelin

Question 14

What was the neural architecture of the agent sponsored by Synapse Core?

(1) State-Space Mamba Hybrid

(2) Diffusion Planner Engine

(3) Autoregressive Transformer

(4) Neuro-Symbolic Reasoner

Question 15

In which of the following statements are the agents correctly listed in descending order of their benchmark rank (highest rank to lowest rank)?

I. Kestrel; the Autoregressive Transformer; the agent sponsored by Apex Systems

II. The Diffusion Planner Engine; the agent sponsored by Synapse Core; Javelin

III. The agent sponsored by Apex Systems; the Autoregressive Transformer; Lynx

(1) I only

(2) II only

(3) III only

(4) I and III

Question 16

Which of the following conclusions can be definitively established from the given benchmark telemetry?

A. Maven is built on the Diffusion Planner Engine architecture.

B. The Neuro-Symbolic Reasoner is agent Lynx.

C. Javelin captures the third rank (Rank 3).

(1) A and B

(2) A and C

(3) A, B and C

(4) B and C

Show solution

Step-by-Step Deductive Solution

1. Parameter Grid & Rank Anchor Analysis

We have four discrete ranks: Rank 1, Rank 2, Rank 3, and Rank 4.

The three attributes to be matched per rank are:

  1. Agent Model Name: Javelin, Kestrel, Lynx, Maven

  2. Neural Architecture: Autoregressive Transformer, Diffusion Planner Engine, State-Space Mamba Hybrid, Neuro-Symbolic Reasoner

  3. Sponsoring Lab: Apex Systems, DeepDynamix, Synapse Core, Radiant Intelligence

From Clues II, III, and V:

  • Rank 1: Architecture = State-Space Mamba Hybrid (Clue II)

  • Rank 2: Agent Model = Maven (Clue III)

  • Rank 4: Sponsoring Lab = DeepDynamix, Architecture = Neuro-Symbolic Reasoner (Clue V)


2. Deducing Javelin, Kestrel, and Lynx

  1. Locating Javelin (Clue IV):

    • Javelin is an Autoregressive Transformer.

    • Javelin cannot be Rank 1 (Mamba Hybrid), Rank 2 (Maven), or Rank 4 (Neuro-Symbolic).

    • Therefore, Javelin MUST be Rank 3, and Rank 3 is built on the Autoregressive Transformer architecture!

  2. Resolving Architectures:

    • Rank 1: State-Space Mamba Hybrid

    • Rank 3: Autoregressive Transformer

    • Rank 4: Neuro-Symbolic Reasoner

    • The sole remaining architecture for Rank 2 (Maven) is the Diffusion Planner Engine!

  3. Locating Kestrel & Lynx (Clues V & VI):

    • Sponsoring lab of Kestrel is Radiant Intelligence (Clue VI).

    • DeepDynamix sponsors Rank 4 (Clue V). Thus, Kestrel cannot be Rank 4.

    • Since Maven is Rank 2 and Javelin is Rank 3, Kestrel MUST be Rank 1!

    • Consequently, Radiant Intelligence sponsors Rank 1 (Kestrel).

    • That leaves Lynx as Rank 4 (sponsored by DeepDynamix, Neuro-Symbolic Reasoner)!

  4. Assigning Remaining Sponsoring Labs (Clue I):

    • Remaining labs: Apex Systems and Synapse Core.

    • Clue I dictates that the agent sponsored by Synapse Core stood neither first nor second.

    • Since Rank 1 is Radiant Intelligence and Rank 2 is open, Synapse Core cannot be Rank 2.

    • Therefore, Synapse Core must sponsor Rank 3 (Javelin)!

    • This leaves Apex Systems to sponsor Rank 2 (Maven)!


Master Agent Benchmark Leaderboard

Benchmark RankAgent ModelNeural ArchitectureSponsoring DeepTech Lab
Rank 1 (First)KestrelState-Space Mamba HybridRadiant Intelligence
Rank 2 (Second)MavenDiffusion Planner EngineApex Systems
Rank 3 (Third)JavelinAutoregressive TransformerSynapse Core
Rank 4 (Fourth)LynxNeuro-Symbolic ReasonerDeepDynamix

Question-by-Question Answers

Question 13:

  • The first rank was captured by Kestrel, sponsored by Radiant Intelligence.

  • Correct Choice: (2) The agent sponsored by Radiant Intelligence

Question 14:

  • Synapse Core sponsored Javelin, which is built on the Autoregressive Transformer architecture.

  • Correct Choice: (3) Autoregressive Transformer

Question 15:

  • Evaluating descending rank order (Rank 1 →\to Rank 2 →\to Rank 3 →\to Rank 4):

    • Statement I: Kestrel (Rank 1); Autoregressive Transformer (Rank 3); Apex Systems (Rank 2). The rank sequence is 1→3→21 \to 3 \to 2, which is NOT descending.

    • Statement II: Diffusion Planner (Rank 2); Synapse Core (Rank 3); Javelin (Rank 3). The sequence is 2→3→32 \to 3 \to 3, which is NOT strictly descending.

    • Statement III: Apex Systems (Rank 2); Autoregressive Transformer (Rank 3); Lynx (Rank 4). The sequence is 2→3→42 \to 3 \to 4, which is strictly descending!

  • Therefore, Statement III only is correct.

  • Correct Choice: (3) III only

Question 16:

  • Conclusion A: Maven is built on the Diffusion Planner Engine (True).

  • Conclusion B: Lynx is the Neuro-Symbolic Reasoner (True).

  • Conclusion C: Javelin captures the third rank (True).

  • Since all three conclusions are definitively established, A, B and C can be drawn.

  • Correct Choice: (3) A, B and C

Solved Example Set 5: Questions 17 to 20 (High-Performance Computing Supercomputer Cluster Audit)

Directions for Questions 17 to 20

A sovereign AI infrastructure auditor evaluated the flagship High-Performance Computing (HPC) supercomputing clusters of five deeptech research consortiums: Aether, Boreas, Chronos, Daedalus, and Eos.

Each consortium's supercomputing cluster belongs to a distinct hardware architecture generation tier: Tier-A, Tier-B, Tier-C, Tier-D, and Tier-E. The architecture tiers correspond strictly to the five chronological deployment epochs in which the clusters were commissioned:

  • Epoch 1: Pre-Volta Era (Prior to 2018) — Oldest

  • Epoch 2: Ampere Architecture Era (2018 to 2020)

  • Epoch 3: Hopper Architecture Era (2021 to 2023)

  • Epoch 4: Blackwell Architecture Era (2024 to 2025)

  • Epoch 5: Photonic/Quantum-Optical Era (2026 onwards) — Newest

Each cluster operates under one of two operational modes:

  • Dedicated Sovereign Private Compute (self-operated by consortium researchers)

  • Commercial Multi-Tenant Hosted Cloud (colocation capacity leased out to external clients)

The auditor's compliance report recorded the following details:

  • Clue I: Chronos's cluster is one of the two clusters operating in Commercial Multi-Tenant Hosted Cloud mode.

  • Clue II: Boreas's supercomputing cluster was commissioned earlier than Aether's cluster.

  • Clue III: Eos's cluster is the newest (Epoch 5) and Daedalus's cluster is the oldest (Epoch 1). Neither of these two clusters belongs to Tier-A or Tier-B.

  • Clue IV: Boreas's cluster belongs to Tier-D.

  • Clue V: The two clusters operating in Commercial Multi-Tenant Hosted Cloud mode belong to Tier-B and Tier-D.

  • Clue VI: Chronos commissioned its cluster in the year 2022 (i.e., strictly within the Hopper Architecture Era: 2021 to 2023).


Question 17

What is the correct chronological sequence of supercomputing clusters in terms of their commissioning epoch (from oldest to newest)?

(1) Daedalus, Boreas, Aether, Chronos, Eos

(2) Daedalus, Chronos, Boreas, Aether, Eos

(3) Daedalus, Boreas, Chronos, Aether, Eos

(4) Daedalus, Aether, Boreas, Chronos, Eos

Question 18

Which of the following combinations of 'Consortium Cluster − Architecture Tier' are necessarily correct?

I. Aether's cluster – Tier-A

II. Daedalus's cluster – Tier-C

III. Chronos's cluster – Tier-B

(1) I and III only

(2) II and III only

(3) III only

(4) All are correct

Question 19

If Eos's cluster belongs to Tier-E, to which architecture tier does Daedalus's cluster belong?

(1) Tier-A

(2) Tier-B

(3) Tier-C

(4) Tier-D

Question 20

Which of the following represents the correct sequence (from newest to oldest) among the clusters operating in Dedicated Sovereign Private Compute mode?

(1) Daedalus, Aether, Eos

(2) Aether, Daedalus, Eos

(3) Eos, Aether, Daedalus

(4) Eos, Daedalus, Aether

Show solution

Step-by-Step Deductive Solution

1. Epochs & Consortium Mapping

Let the 5 deployment epochs be ordered chronologically from 1 (Oldest) to 5 (Newest):

  • Epoch 1: Prior to 2018 (Oldest)

  • Epoch 2: 2018 to 2020

  • Epoch 3: 2021 to 2023 (Hopper Era)

  • Epoch 4: 2024 to 2025 (Blackwell Era)

  • Epoch 5: 2026 onwards (Newest)

From Clue III and Clue VI:

  • Daedalus is the oldest   ⟹  \implies Daedalus = Epoch 1 (Prior to 2018).

  • Eos is the newest   ⟹  \implies Eos = Epoch 5 (2026 onwards).

  • Chronos commissioned its cluster in 2022   ⟹  \implies Chronos = Epoch 3 (2021 to 2023).

This leaves Epoch 2 and Epoch 4 for Boreas and Aether.

  • Clue II states that Boreas's cluster was commissioned earlier than Aether's cluster.

  • Therefore:

    • Boreas = Epoch 2 (2018 to 2020)

    • Aether = Epoch 4 (2024 to 2025)


2. Deducing Operational Modes & Architecture Tiers

  1. Operational Modes (Clues I & V):

    • Exactly two clusters operate in Commercial Multi-Tenant Hosted Cloud mode.

    • Clue V states that the two multi-tenant hosted clusters belong to Tier-B and Tier-D.

    • Clue I states that Chronos is multi-tenant.

    • Clue IV states that Boreas belongs to Tier-D.

    • Thus, Boreas is one of the multi-tenant hosted clusters (Tier-D).

    • Therefore, Chronos must be the other multi-tenant hosted cluster, which belongs to Tier-B!

    • The remaining three consortiums (Daedalus, Aether, Eos) operate in Dedicated Sovereign Private Compute mode.

  2. Resolving Architecture Tiers:

    • We have:

      • Chronos (Epoch 3) = Tier-B

      • Boreas (Epoch 2) = Tier-D

    • The remaining tiers are Tier-A, Tier-C, and Tier-E, to be assigned among Daedalus (Epoch 1), Aether (Epoch 4), and Eos (Epoch 5).

    • Clue III states that neither Daedalus nor Eos belongs to Tier-A or Tier-B.

    • Since Tier-A cannot belong to Daedalus or Eos, Tier-A MUST belong to Aether (Epoch 4)!

    • Daedalus and Eos divide the remaining two tiers: Tier-C and Tier-E.

      • If Eos is Tier-E, Daedalus is Tier-C.

      • If Eos is Tier-C, Daedalus is Tier-E.


Master Supercomputer Cluster Audit Table

ConsortiumChronological Commissioning EpochHardware Generation TierOperational Deployment Mode
DaedalusEpoch 1: Prior to 2018 (Oldest)Tier-C or Tier-EDedicated Sovereign Private Compute
BoreasEpoch 2: 2018 to 2020Tier-DCommercial Multi-Tenant Hosted Cloud
ChronosEpoch 3: 2021 to 2023 (Hopper)Tier-BCommercial Multi-Tenant Hosted Cloud
AetherEpoch 4: 2024 to 2025 (Blackwell)Tier-ADedicated Sovereign Private Compute
EosEpoch 5: 2026 onwards (Newest)Tier-E or Tier-CDedicated Sovereign Private Compute

Question-by-Question Answers

Question 17:

  • The chronological sequence from oldest to newest is:

    Epoch 1 (Daedalus) →\to Epoch 2 (Boreas) →\to Epoch 3 (Chronos) →\to Epoch 4 (Aether) →\to Epoch 5 (Eos).

  • Correct Choice: (3) Daedalus, Boreas, Chronos, Aether, Eos

Question 18:

  • Statement I: Aether's cluster – Tier-A (Necessarily Correct).

  • Statement II: Daedalus's cluster – Tier-C (Not necessarily correct; could be Tier-E).

  • Statement III: Chronos's cluster – Tier-B (Necessarily Correct).

  • Thus, only I and III are necessarily correct.

  • Correct Choice: (1) I and III only

Question 19:

  • Daedalus and Eos divide Tier-C and Tier-E. If Eos's cluster belongs to Tier-E, Daedalus's cluster must belong to Tier-C.

  • Correct Choice: (3) Tier-C

Question 20:

  • The clusters operating in Dedicated Sovereign Private Compute mode are Eos (Epoch 5), Aether (Epoch 4), and Daedalus (Epoch 1).

  • Ordered from newest to oldest: Eos, Aether, Daedalus.

  • Correct Choice: (3) Eos, Aether, Daedalus

Practice questions

Question 1 · Foundation

Diagnostic Set: Propositional Selection Logic

A frontier AI safety research institute must select a 5-member red-teaming squad from a pool of 8 elite researchers: A, B, C, D, E, F, G, and H.

The selection committee enforces the following operational constraints:

  1. Exactly one of A or B must be selected (A⊕BA \oplus B).

  2. If C is selected, then D must also be selected (C  ⟹  DC \implies D).

  3. E and F are mutually exclusive; they cannot both be selected for the squad.

  4. G can be selected if and only if H is selected (G  ⟺  HG \iff H).

  5. If B is selected, then E must be selected (B  ⟹  EB \implies E).

If Researcher C is selected as a member of the squad, who among the following CANNOT be selected in the squad?

  1. A

    Researcher A

  2. B

    Researcher D

  3. C

    Researcher E

  4. D

    Researcher G

Show answer and explanation

Answer: (C) Researcher E

Step-by-Step Deductive Proof:

  1. Given Condition:

    • CC is in the squad (xC=1x_C = 1).

    • Rule 2 states C  ⟹  DC \implies D, so DD must also be selected (xD=1x_D = 1).

    • Candidates confirmed so far: {C,D}\{C, D\}. Squad requires 3 more members.

  2. Evaluating the Pair {G, H} (Rule 4):

    • GG and HH are a tied package deal (G  ⟺  HG \iff H). They are either both selected or neither selected.

    • Case 1: Both G and H are selected.

      • Confirmed members: {C,D,G,H}\{C, D, G, H\} (4 members).

      • We need exactly 1 more member from {A,B,E,F}\{A, B, E, F\}.

      • Rule 1 requires exactly one of AA or BB to be in the squad.

      • If BB is chosen: Rule 5 dictates B  ⟹  EB \implies E, which would require EE to be chosen too, creating a 6-member squad! Contradiction.

      • Hence, BB cannot be chosen. Therefore, AA MUST be chosen.

      • Since AA is the 5th member, no further candidates can be chosen.

      • Thus, the squad is uniquely: {A,C,D,G,H}\{A, C, D, G, H\}.

    • Case 2: Neither G nor H is selected.

      • We have {C,D}\{C, D\} and need 3 more members from {A,B,E,F}\{A, B, E, F\}.

      • From Rule 1, exactly one of AA or BB is chosen.

      • That leaves 2 spots to fill from the remaining 2 candidates {E,F}\{E, F\}.

      • But Rule 3 strictly forbids selecting both EE and FF! Contradiction.

      • Hence, Case 2 is completely impossible.

  3. Conclusion:

    • The only valid squad containing CC is {A,C,D,G,H}\{A, C, D, G, H\}.

    • Neither B,E,B, E, nor FF can be in the squad.

    • Among the given options, Researcher E CANNOT be selected.

Question 2 · Foundation

Diagnostic Set: Capacity-Constrained Partitioning

A DeepTech incubator distributes 21 autonomous AI engineering interns across 4 project divisions: Robotics, NLP, Vision, and Quantum.

The distribution must satisfy the following integer capacity bounds:

  1. Each division must receive at least 3 interns, and no division can receive more than 8 interns (3≤∣Gj∣≤83 \le |G_j| \le 8).

  2. The Robotics division must receive strictly more interns than the NLP division (R>NR > N).

  3. The Vision division must receive strictly more interns than the Quantum division (V>QV > Q).

  4. The number of interns assigned to the Quantum division is an even number.

What is the maximum possible number of interns that can be assigned to the NLP division?

  1. A

    4

  2. B

    5

  3. C

    6

  4. D

    7

Show answer and explanation

Answer: (B) 5

Step-by-Step Mathematical Derivation:

  1. Integer Constraints:

    • R+N+V+Q=21R + N + V + Q = 21

    • 3≤R,N,V,Q≤83 \le R, N, V, Q \le 8

    • R≥N+1R \ge N + 1 (since R>NR > N)

    • V≥Q+1V \ge Q + 1 (since V>QV > Q)

    • Q∈{4,6,8}Q \in \{4, 6, 8\} (even and ≥3\ge 3)

  2. Bounding the Vision-Quantum Subsystem (V+QV + Q):

    • Since QQ is even and V>QV > Q with V≤8V \le 8:

      • If Q=8Q = 8, V>8V > 8, impossible since max is 8.

      • If Q=6Q = 6, V∈{7,8}  ⟹  V+Q∈{13,14}V \in \{7, 8\} \implies V + Q \in \{13, 14\}.

      • If Q=4Q = 4, V∈{5,6,7,8}  ⟹  V+Q∈{9,10,11,12}V \in \{5, 6, 7, 8\} \implies V + Q \in \{9, 10, 11, 12\}.

    • The absolute minimum sum for (V+Q)(V + Q) is 5+4=95 + 4 = 9.

  3. Maximizing NN (NLP Division):

    • To maximize NN, we must minimize (V+Q)(V + Q) and keep RR as small as possible relative to NN:

      R+N=21−(V+Q)R + N = 21 - (V + Q)

    • Since (V+Q)≥9(V + Q) \ge 9, we have:

      R+N≤21−9=12R + N \le 21 - 9 = 12

    • Since R>NR > N, we must have R≥N+1R \ge N + 1:

      (N+1)+N≤R+N≤12  ⟹  2N+1≤12  ⟹  2N≤11  ⟹  N≤5.5(N + 1) + N \le R + N \le 12 \implies 2N + 1 \le 12 \implies 2N \le 11 \implies N \le 5.5

    • Since NN is an integer, N≤5N \le 5.

  4. Feasibility Verification for N=5N = 5:

    • Set N=5N = 5.

    • Since R>NR > N, minimal R=6R = 6.

    • Then R+N=6+5=11R + N = 6 + 5 = 11.

    • Remaining sum: V+Q=21−11=10V + Q = 21 - 11 = 10.

    • Set Q=4Q = 4 (even) and V=6V = 6 (V>QV > Q and V≤8V \le 8).

    • Intern allocation: R=6,N=5,V=6,Q=4R = 6, N = 5, V = 6, Q = 4.

    • Sum =6+5+6+4=21= 6 + 5 + 6 + 4 = 21. All constraints satisfied!

  5. Therefore, the maximum number of interns that can be assigned to the NLP division is 5.

Question 3 · Foundation

Diagnostic Set: Multi-Attribute Transitive Matching

Four DeepTech researchers—Kiran, Liam, Mira, and Neil—each specialize in a distinct AI field: Computer Vision, NLP, Reinforcement Learning (RL), and Robotics.

Each researcher operates from a distinct global innovation hub: San Francisco, London, Tokyo, and Zurich, and each is assigned to lead one of four compute clusters: Cluster-1, Cluster-2, Cluster-3, and Cluster-4.

The following facts are confirmed:

  1. The researcher operating from Tokyo specializes in Computer Vision, but is not assigned to Cluster-1.

  2. Kiran specializes in Robotics and is assigned to Cluster-3.

  3. Liam operates from San Francisco, but does not specialize in NLP.

  4. The researcher assigned to Cluster-4 specializes in Reinforcement Learning (RL) and does not operate from Zurich.

Which compute cluster is assigned to the researcher who specializes in Computer Vision?

  1. A

    Cluster-1

  2. B

    Cluster-2

  3. C

    Cluster-3

  4. D

    Cluster-4

Show answer and explanation

Answer: (B) Cluster-2

Step-by-Step Deductive Matching:

  1. Attribute Linkages from Direct Clues:

    • Clue 1: Tokyo ↔\leftrightarrow Computer Vision (CV), Cluster ≠\ne Cluster-1.

    • Clue 2: Kiran ↔\leftrightarrow Robotics ↔\leftrightarrow Cluster-3.

    • Clue 4: Cluster-4 ↔\leftrightarrow Reinforcement Learning (RL), Hub ≠\ne Zurich.

  2. Identifying Liam's Specialization & Cluster:

    • Liam is from San Francisco (Clue 3).

    • Liam cannot specialize in CV (Tokyo is CV), Robotics (Kiran is Robotics), or NLP (Clue 3).

    • Therefore, Liam MUST specialize in Reinforcement Learning (RL)!

    • From Clue 4, the RL specialist is assigned to Cluster-4.

    • Thus: Liam ↔\leftrightarrow San Francisco ↔\leftrightarrow RL ↔\leftrightarrow Cluster-4!

  3. Locating the Computer Vision Specialist's Cluster:

    • The four clusters are Cluster-1, Cluster-2, Cluster-3, and Cluster-4.

    • Cluster-3 is assigned to Kiran (Robotics).

    • Cluster-4 is assigned to Liam (RL).

    • The Computer Vision specialist (who is based in Tokyo) cannot be assigned to Cluster-1 (Clue 1).

    • Therefore, the Computer Vision specialist MUST be assigned to Cluster-2!

    • Consequently, the NLP specialist must be assigned to Cluster-1.

  4. Conclusion:

    • The researcher specializing in Computer Vision is definitively assigned to Cluster-2.

Question 4 · Foundation

Diagnostic Set: System Integration Module Selection

An autonomous drone fleet development project requires selecting a core architecture of exactly 4 software modules from a candidate set of 7 modules: Nav, Vision, Sensor, Control, Comm, Power, and Telemetry.

The integration pipeline is governed by the following system architecture rules:

  1. If Nav is selected, then Vision cannot be selected (Nav  ⟹  ¬VisionNav \implies \neg Vision).

  2. Sensor and Control must either both be selected or neither can be selected (Sensor  ⟺  ControlSensor \iff Control).

  3. If Comm is selected, then Nav must be selected (Comm  ⟹  NavComm \implies Nav).

  4. If Vision is not selected, then Power must be selected (¬Vision  ⟹  Power\neg Vision \implies Power).

  5. If Telemetry is selected, then Control cannot be selected (Telemetry  ⟹  ¬ControlTelemetry \implies \neg Control).

If the Comm module is selected as one of the 4 modules, which of the following is the complete and exact set of the other three selected modules?

  1. A

    Nav, Sensor, and Control

  2. B

    Nav, Power, and Telemetry

  3. C

    Power, Sensor, and Telemetry

  4. D

    Vision, Power, and Control

Show answer and explanation

Answer: (B) Nav, Power, and Telemetry

Step-by-Step Chain of Deductions:

  1. Given Assumption:

    • Comm is selected.
  2. Triggering Implication Chains:

    • Rule 3: Comm  ⟹  NavComm \implies Nav. Since Comm is selected, Nav MUST be selected.

    • Rule 1: Nav  ⟹  ¬VisionNav \implies \neg Vision. Since Nav is selected, Vision CANNOT be selected.

    • Rule 4: ¬Vision  ⟹  Power\neg Vision \implies Power. Since Vision is not selected, Power MUST be selected.

    • Selected so far: {Comm, Nav, Power} (3 modules). Exactly 1 more module is needed.

  3. Evaluating Remaining Candidates {Sensor, Control, Telemetry}:

    • Rule 2 dictates that Sensor and Control must be selected together or neither.

    • Since only 1 module slot remains in the 4-module architecture, we cannot select both Sensor and Control!

    • Therefore, neither Sensor nor Control can be selected.

    • The sole remaining candidate to fill the 4th slot is Telemetry.

    • Check Rule 5: Since Control is not selected, Telemetry is fully valid.

  4. Final Architecture:

    • The complete selected architecture is: {Comm, Nav, Power, Telemetry}.

    • The other three selected modules are Nav, Power, and Telemetry.

Question 5 · Foundation

Diagnostic Set: Categorical Distribution & Pigeonhole Allocation

A venture capital syndicate distributes 10 portfolio AI startups into three distinct investment cohorts: Seed, Series A, and Series B.

  1. Every cohort must contain at least 2 startups.

  2. The number of startups in each cohort is strictly distinct (no two cohorts contain the same number of startups).

  3. The Series A cohort contains strictly more startups than the Seed cohort (∣Series A∣>∣Seed∣|\text{Series A}| > |\text{Seed}|).

  4. Exactly 6 of the 10 startups are Generative AI startups, and every cohort contains at least one Generative AI startup.

What is the maximum possible number of Generative AI startups that can be allocated to the Seed cohort?

  1. A

    2

  2. B

    3

  3. C

    4

  4. D

    5

Show answer and explanation

Answer: (B) 3

Step-by-Step Combinatorial Derivation:

  1. Determining Valid Integer Partitions of 10:

    • Let the cohort sizes be SS (Seed), AA (Series A), and BB (Series B).

    • S+A+B=10S + A + B = 10, with S,A,B≥2S, A, B \ge 2 and all three distinct.

    • A>SA > S.

    • Testing distinct positive integer triples summing to 10 with minimum part ≥2\ge 2:

      • The only unordered set of three distinct integers summing to 10 with minimum ≥2\ge 2 is:

        {2,3,5}(2+3+5=10)\{2, 3, 5\} \quad (2 + 3 + 5 = 10)

      • Note: {2,4,4}\{2, 4, 4\} fails distinctness; {1,… }\{1, \dots\} fails the minimum threshold of 2.

    • Since A>SA > S, the permissible ordered assignments for (S,A,B)(S, A, B) are:

      • Case 1: S=2,A=3,B=5S = 2, A = 3, B = 5

      • Case 2: S=2,A=5,B=3S = 2, A = 5, B = 3

      • Case 3: S=3,A=5,B=2S = 3, A = 5, B = 2

  2. Allocating Generative AI Startups:

    • Total Generative AI startups =6= 6.

    • Let gS,gA,gBg_S, g_A, g_B be the number of Generative AI startups in Seed, Series A, and Series B.

    • gS+gA+gB=6g_S + g_A + g_B = 6.

    • Boundary constraints:

      • gS≤S,gA≤A,gB≤Bg_S \le S, \quad g_A \le A, \quad g_B \le B

      • gS≥1,gA≥1,gB≥1g_S \ge 1, \quad g_A \ge 1, \quad g_B \ge 1

    • To maximize gSg_S, we minimize gAg_A and gBg_B:

      gS=6−(gA+gB)≤6−(1+1)=4g_S = 6 - (g_A + g_B) \le 6 - (1 + 1) = 4

    • However, gSg_S cannot exceed the cohort capacity SS (gS≤Sg_S \le S).

      • In Case 1 and Case 2, S=2  ⟹  gS≤2S = 2 \implies g_S \le 2.

      • In Case 3, S=3  ⟹  gS≤3S = 3 \implies g_S \le 3.

  3. Verifying Feasibility of gS=3g_S = 3 (Case 3):

    • Cohort sizes: S=3,A=5,B=2S = 3, A = 5, B = 2.

    • Assign gS=3g_S = 3 (all 3 startups in Seed are GenAI).

    • We need gA+gB=6−3=3g_A + g_B = 6 - 3 = 3.

    • We can set gA=2g_A = 2 (2≤52 \le 5) and gB=1g_B = 1 (1≤21 \le 2).

    • Every cohort contains at least one GenAI startup (3≥1,2≥1,1≥13 \ge 1, 2 \ge 1, 1 \ge 1).

    • Sum =3+2+1=6= 3 + 2 + 1 = 6. All conditions are satisfied!

  4. Conclusion:

    • The maximum possible number of Generative AI startups in the Seed cohort is 3.

Try real CAT questions

Study this concept with interactive practice in PrepSoc.

Open the concept library