llm-training-alignment

36 entries in Machine Learning & AI.

• Tensor parallelismRow-column weight shardingstandardllm-training-alignment
• Pipeline parallelismMicro-batch bubble reductionstandardllm-training-alignment
• Sequence parallelismActivation sharding along tokensspecialistllm-training-alignment
• Expert parallelismToken routing across devicesspecialistllm-training-alignment
• ZeROOptimizer-state shardingstandardllm-training-alignment
• Fully sharded data parallelShard-gather-reshard parametersstandardllm-training-alignment
• Gradient checkpointingRecompute activations on backwardstandardllm-training-alignment
• Gradient accumulationMicro-batch gradient summationstandardllm-training-alignment
• Mixed-precision trainingLoss scalingstandardllm-training-alignment
• FP8 trainingPer-tensor scaling factorsspecialistllm-training-alignment
• QLoRAFour-bit NormalFloat base weightsstandardllm-training-alignment
• Prefix tuningLearned prefix key-value vectorsspecialistllm-training-alignment
• RLHFProximal policy optimization on a reward modelcanonllm-training-alignment
• Direct preference optimizationImplicit reward reparameterizationcanonllm-training-alignment
• IPOIdentity preference objectivespecialistllm-training-alignment
• KTOProspect-theoretic utility lossspecialistllm-training-alignment
• ORPOOdds-ratio preference lossspecialistllm-training-alignment
• GRPOGroup-relative advantage baselinestandardllm-training-alignment
• Constitutional AISelf-critique and revisionstandardllm-training-alignment
• Rejection sampling fine-tuningBest-of-n distillationspecialistllm-training-alignment
• Reward model trainingBradley-Terry pairwise lossstandardllm-training-alignment
• Process reward modelingStep-level supervisionspecialistllm-training-alignment
• Chain-of-thought promptingStep-by-step elicitationcanonllm-training-alignment
• Self-consistencyMajority vote over sampled chainsstandardllm-training-alignment
• Tree of thoughtsDeliberate search over partial thoughtsstandardllm-training-alignment
• Least-to-most promptingSubproblem decompositionspecialistllm-training-alignment
• ReActInterleaved reasoning and actingcanonllm-training-alignment
• ToolformerSelf-supervised API call insertionspecialistllm-training-alignment
• ReflexionVerbal self-feedback memoryspecialistllm-training-alignment
• LLM-as-judgePairwise preference scoringstandardllm-training-alignment
• Perplexity evaluationHeld-out token likelihoodstandardllm-training-alignment
• Pass-at-k estimationUnbiased sampling estimatorspecialistllm-training-alignment
• Sequence-level distillationTeacher-generated targetsspecialistllm-training-alignment
• Curriculum data mixingPhase-scheduled corpus weightsspecialistllm-training-alignment
• Deduplication for pretrainingSuffix-array exact substring removalspecialistllm-training-alignment
• Data quality filteringClassifier-scored web textspecialistllm-training-alignment