Skip to content
tezvyn:

All bites

The whole library, newest first. Filter by what you are here for, or pick a topic if you already know.

4330 bites

Page 47

LLMs & Generative AI1 min read

Measuring fairness in an embedding-based text classifier

Define group definitions, apply demographic parity and equalized odds, and address noisy labels and implicit group membership.

LLMs & Generative AI1 min read

Practical explainability for an LLM loan summary

Use attribution-by-design with grounded citations, structured rationales, and a deterministic rules layer instead of slow per-token SHAP.

LLMs & Generative AI1 min read

Data lineage and machine unlearning for a fine-tuned LLM

Version and fingerprint datasets, record transforms and which checkpoint saw what, and enable unlearning via retraining, data sharding, or approximate gradient methods.

LLMs & Generative AI2 min read

Designing an autonomous research-and-report agent

Planner that decomposes goals, short-term scratchpad plus long-term vector memory, structured tool calls, and a reflect-retry loop for error correction.

LLMs & Generative AI2 min read

Why RAG persists despite million-token context windows

Cost and latency scale with context, attention degrades in the middle, and RAG adds freshness, access control, and citations.

LLMs & Generative AI2 min read

Self-attention and the Query, Key, Value matrices

Queries score against keys via scaled dot product, softmax yields weights, and those weight the values into the output.

LLMs & Generative AI1 min read

How Transformers encode token position

Attention is permutation-invariant, so positional encodings (sinusoidal, learned, or rotary) are added or applied.

LLMs & Generative AI1 min read

Cross-attention versus self-attention in encoder-decoder Transformers

Cross-attention draws Queries from the decoder and Keys/Values from the encoder, letting the decoder condition on the source.

LLMs & Generative AI2 min read

Tokens and vocabulary-size tradeoffs

A token is a subword unit; larger vocab shortens sequences but bloats the embedding matrix, smaller vocab generalizes but lengthens sequences.

LLMs & Generative AI2 min read

Fault-tolerant checkpointing for thousand-GPU pre-training

Checkpoint weights, optimizer state, RNG, and data position together; use asynchronous sharded writes and automated detect-restart-resume.

LLMs & Generative AI2 min read

Prompt engineering to curb extraction hallucinations

Ground strictly in source, allow null for missing fields, enforce a schema, and use few-shot examples; acknowledge prompting cannot fully eliminate it.

LLMs & Generative AI2 min read

Self-consistency over chain-of-thought

Sample multiple CoT paths at nonzero temperature and majority-vote the final answer; cost scales with the number of samples.

LLMs & Generative AI2 min read

Why chain-of-thought helps large models but not small ones

Small models lack reliable multi-step reasoning, so CoT just adds error-prone steps; adapt by using few-shot/fine-tuning or distillation for small tiers.

LLMs & Generative AI2 min read

Handling a 401 error in an LLM agent's tool call

Catch the tool error, return a structured observation to the LLM, and distinguish recoverable retries from terminal failures needing re-plan or escalation.

LLMs & Generative AI2 min read

CLIP's contrastive objective and zero-shot classification

Train image and text encoders to align matched pairs and repel mismatched ones in a shared space; classify zero-shot by comparing an image to text prompts of class names.

LLMs & Generative AI2 min read

Evaluating faithfulness and compositionality in multimodal models

Use targeted probes with hard negatives, attribute-relation binding tests, and structured grounding checks; note BLEU rewards surface overlap not correctness.

LLMs & Generative AI2 min read

Red teaming a generative model

Deliberately probe for harmful outputs across categories, document jailbreaks, and automate with adversarial prompt generators plus classifier-based judging.

LLMs & Generative AI2 min read

Preventing PII in LLM outputs: curation, fine-tuning, or guardrails

Favor post-processing guardrails as the enforceable last line, backed by data curation; note each layer's tradeoffs and that defense-in-depth is best.

LLMs & Generative AI1 min read

Constitutional AI versus standard RLHF

A written principle set guides self-critique and revision, plus AI feedback (RLAIF) replaces human preference labels.

LLMs & Generative AI1 min read

Penalizing sycophancy in a reward model

Sycophancy is reward proxy gaming where agreeableness substitutes for correctness; counter it with truth-anchored labels, perturbed-premise pairs, and consistency checks.