← Back
Home › Systems › Frontier AI research

safety-classifiers

misuse classifiers on model inputs and outputs

constitutional classifiers

SystemTechniques · 8Primitives · 39 safety-classifiers Active learningmachine-learning · medium: Choosing which unlabeled examples to label next: uncertainty, margin, diversity and committee samplingChunked streaming attentionmachine-learning · medium: Chunk-limited attention masks with cached left context so encoders (Conformer, Whisper-style) run streaming at bounded latencyDistribution drift detectionmachine-learning · medium: PSI, KS and other tests for feature and score drift between training and productionLLM-as-judgemachine-learning · medium: Grade open-ended outputs with a language model under a rubric, then correct its known biases by order swapping, voting and calibration against human labels.›Text classificationmachine-learning · medium: Assign one or more labels to a text; climb a ladder from regexes to linear models on n-grams, embeddings, fine-tuned encoders and LLM prompts, only as far as the data and budget demand.›Dynamic batchingml-training · medium: Queue independent requests and dispatch them as one batch when it is full or the oldest has waited max_delay; bounded extra latency buys throughput.›Knowledge distillationml-training · medium: Train a small student to match a large teacher's softened outputs (plus the hard labels) so it keeps most of the teacher's accuracy at a fraction of the cost.›Calibrationmachine-learning · low: Make predicted probabilities match observed frequencies: measure with ECE, fix post hoc with temperature, Platt or isotonic scaling on held-out data.›Transformer blockml-training · low: The decoder-only pre-norm block (attention plus MLP on a residual stream) behind every modern LLM, with exact shapes, parameter, FLOP and KV-cache counts.›Annotator agreementstatistics · low: Inter-annotator agreement measures: Cohen and Fleiss kappa, Krippendorff alpha, gold-question accuracyData augmentationmachine-learning · low: Generate label-preserving variants of each training example on the fly to enlarge the effective dataset and bake in invariances.›Loss functionsmachine-learning · low: The scalar a model minimizes: cross-entropy for classes and tokens, MSE or Huber for values, each with a stable form and simple gradient.›Prompt cachingagentic-ai · low: Reuse the attention keys/values of a repeated prompt prefix, so an agent's growing context is prefilled once and later turns pay only for the new tokens.›Back translationmachine-learning · low: Data augmentation for translation: translate monolingual target text back to source to make synthetic parallel pairsEvaluation metricsmachine-learning · low: Precision, recall, ROC AUC, average precision, log loss, NDCG and pass@k computed correctly, plus paired confidence intervals for comparing models.›Template-based text generationmachine-learning · low: Realise facts as sentences by filling slots in alternative templates, with agreement, referring expressions and seeded variation; every word traces to data or a template.›Lora adaptersml-training · low: Low-rank adapter matrices for fine-tuning; QLoRA; serving many adapters on one base modelLabel aggregationstatistics · low: Combining noisy labels from several annotators: majority vote, Dawid-Skene EM, annotator weightingEmbeddingsmachine-learning · low: A learned table mapping discrete ids to dense vectors: gather forward, sparse scatter-add backward, row-wise optimizers to keep huge tables cheap.›k-means clusteringmachine-learning · low: Partition points into k clusters by alternating nearest-centroid assignment and centroid averaging; k-means++ seeding makes it dependable.›Attention variantsmachine-learning · low: Multi-head, multi-query, grouped-query, sliding-window and latent attention; same softmax(QKᵀ/√d)V core, different trade-offs between quality and KV-cache size.›Quantile sketchesdata-structures · low: Mergeable quantile sketches: t-digest, DDSketch, HDR histogramsToken samplingml-training · low: Turn a vector of scores into one chosen item with temperature, top-k, top-p, min-p, repetition penalties and a seed.›+ 1 more techniques+ 23 more primitives

Similar systems

Share the most building blocks with this one.

prompt-injection-detection 36 %abuse-classifier 27 %model-compression 16 %