Adobe
Ranking, recommendations and LLM serving. The heaviest reported topics are transformer attention, KV-cache trade-offs, and NDCG-style ranking metrics.
26 problems tagged for Adobe.
Sign in to track your progress0%
Solve a problem (pass its tests) and it's checked off here automatically — your progress is saved, so it stays checked when you come back.
- Cross-attention — decoder Q, encoder K/VMedium
- Map a model to its transformer familyEasy
- Project X into Q, K, VEasy
- Full self-attention from XMedium
- Which transformer family?Easy
- Causal mask: build + applyEasy
- Transformer block forward (pre-LN, residual)Hard
- ReLU forward + backwardEasy
- Sigmoid forward + backwardEasy
- Cross-Attention (Encoder-Decoder)Medium
- Expected Reciprocal Rank (ERR)Medium
- The autoregressive generation loopMedium
- Why mask before softmax, not afterMedium
- The residual Jacobian I + ∂f/∂xEasy
- K-means: one iterationMedium
- KV cache for autoregressive inferenceMedium
- KV cache compression (MLA)Hard
- LambdaMART Lambda GradientsHard
- Negative Sampling (Implicit Feedback)Medium
- Neural ODE forward EulerMedium
- Pairwise cosine-similarity matrixEasy
- Pairwise Ranking Loss (BPR / RankNet)Medium
- Precision@k and NDCG@kMedium
- Residual block with shortcutEasy
- Scaled dot-product attentionMedium
- Two-Tower Top-k RetrievalEasy