Short Answer

Picking a cheap but predictive eval to track during training: what do you actually log?

Why has GRPO eaten PPO's lunch in 2025-2026 reasoning post-training?

OpenAI's hosted FT for tool-calling models: what does a JSONL row need beyond text SFT?

Walk through one training example for SFT'ing a single weather-tool call.

Replay-buffer mix for domain SFT: how much general data, and from where?

Designing a code-assistant SFT mix: how should domain and general data balance?

Design a fine-tune that teaches a model to refuse when the retrieved context lacks the answer.

Pretraining vs fine-tuning: two key differences in data and objective.

DeepSeek-style reasoning distillation: what travels from teacher to student, and what gets cut from the loss?

Loss spikes the moment warmup ends: which hyperparameter is the prime suspect?

A DPO run spikes to NaN around step 400 after a clean start: explain the most likely cause

Train loss drops while eval loss climbs at step 1200: name it and the standard fixes

Contrast SFT and continued pretraining when adapting a base model to medical text

Compare LoRA r=8 vs r=64 on a 13B model on the quality-vs-cost curve

Describe the JSON shape of a Claude Opus 4.7 assistant message that calls a tool

Defend the claim that full fine-tuning still beats LoRA for some 2026 production workloads

How does a Vision Transformer turn an image into a sequence, and what mask runs over it?

Define T5's relative position bias and what bucketing buys you

Walk through KV-cache updates when speculative decoding rejects some draft tokens

Explain effective context for Mistral's 4096 sliding window over a 16k input

Sliding-window attention can't see beyond the window. Describe the trick that recovers older context.

Why not just retrain models to eliminate attention sinks entirely?

Why does normalizing Q and K before the dot product help training at scale?

Beyond shape preservation, name two things W_O does after head concat

Why does MQA underperform GQA when both share KV across heads?