Free practice problems
Free LLMs & GenAI Interview Questions
Transformers, attention, tokenization, fine-tuning, RLHF, RAG, evals, inference. 59 of the 320 LLMs & GenAI problems in the LeetML bank are free (3 easy, 36 medium, 20 hard). Open any of them to read the statement, then solve it in your browser.
59 free LLMs & GenAI problems
- What the KV Cache Stores and Why It Speeds Up DecodingConceptualMediumSolve
- How Self-Attention Works, From Intuition to Tensor ShapesConceptualMediumSolve
- Scaled Dot-Product AttentionCodingMediumSolve
- Transformer Attention: Scaling, Heads and MemoryConceptualMediumSolve
- Transformer Encoder LayerCodingHardSolve
- Roofline Estimate: 70B Decoding Speed on One GPUEstimationMediumSolve
- Debug a Left-Padded Causal LM LossCodingHardSolve
- Tokenizers and the Modern LLM BlockConceptualMediumSolve
- LLM Fundamentals: Attention Cost, Causal Masks and RLHFConceptualMediumSolve
- Why LLMs Hallucinate: Diagnosing Four Failure PatternsConceptualMediumSolve
- Self-Attention and Cross-Entropy From ScratchConceptualMediumSolve
- Why Long Context Is Hard for TransformersConceptualHardSolve
- Fine-Tuning a Language Model on Question-Answer PairsConceptualMediumSolve
- Transformer Attention and the 1/sqrt(d_k) ScalingConceptualMediumSolve
- LLM Serving: Latency vs Throughput and Batched InferenceConceptualMediumSolve
- Debugging a RAG Chatbot That Is Confidently WrongConceptualMediumSolve
- GPU Memory Budget for Serving a 70B ModelEstimationMediumSolve
- Reward Model Loss from Pairwise PreferencesCodingMediumSolve
- RLOO: REINFORCE Leave-One-Out AdvantagesCodingMediumSolve
- Causal and Padding Masks in Batched Decoder AttentionConceptualMediumSolve
- Chunking Strategies for RAGConceptualMediumSolve
- Annotation Agreement and LLM JudgesConceptualMediumSolve
- Tool-Calling Weather Agent LoopCodingMediumSolve
- Q/K/V Calculations and FLOPs in Transformer AttentionCalculationMediumSolve
- Learned vs Sinusoidal Positional EncodingsConceptualMediumSolve
- RAG vs Fine-Tuning for Knowledge AssistantsConceptualMediumSolve
- Parameter Count and Attention Share of a 64-Layer TransformerCalculationMediumSolve
- Swap a Text Game's Heuristics for LLM CallsDesignMediumSolve
- Elo Ratings for Model ComparisonCodingMediumSolve
- Chunked Prefill Scheduling Alongside DecodeCodingMediumSolve
- Estimating KV-Cache Memory and Shrinking ItCalculationMediumSolve
261 more LLMs & GenAI problems are in the full bank inside the app.