全部标签

adaptive-computation 4 agent-harness 1 agent-memory 2 agentic-ai 1 agents 2 algebraic-complexity 1 algorithms 1 associative-memory 11 associative-recall 15 attention 26 attention-sinks 1 attribution 1 automata 1 automated-research 2 autoregressive-models 1 bayesian-filtering 3 brain-encoding 1 byte-models 1 capacity-scheduling 1 CCA 1 chain-of-thought 1 CKA 1 communication-complexity 1 complexity 1 compression 2 concept-models 2 conditional-computation 1 consistency-models 1 context-compaction 1 context-distillation 1 context-management 3 context-ranking 1 continual-learning 5 continuous-diffusion 2 copy-mechanism 1 covariance 1 credit-assignment 1 cross-entropy 1 cross-layer-routing 5 cross-layer-sharing 1 curriculum-learning 1 data-augmentation 1 data-distribution 1 decision-models 1 delta-rule 4 depth-scaling 1 diffusion 2 diffusion-language-models 8 distillation 20 domain-adaptation 1 dynamic-routing 3 efficiency 1 efficient-inference 45 efficient-training 6 efficient-transformer 1 embedding-gating 1 embedding-memory 1 embeddings 1 ensemble 1 exact-topk 1 expressivity 5 factual-knowledge 1 fast-weights 6 fine-tuning 3 finite-scalar-quantization 1 fixed-point 1 flow-matching 4 fmri 1 gated-delta-net 1 gated-deltanet 1 generalization 2 generative-replay 1 gpu 1 gpu-kernels 5 gradient-bottleneck 1 gradient-compression 1 graph 1 graph-diffusion 1 grokking 1 hashing 4 hybrid-architectures 15 hybrid-attention 1 hyper-connections 1 identifiability 1 in-context-learning 6 inference 11 interpretability 1 knowledge-editing 2 kv-cache 39 language-model 1 language-modeling 12 language-models 3 latent-language-models 2 latent-reasoning 3 learned-order 1 learning-theory 2 length-generalization 5 linear-attention 64 linear-probes 1 llm 1 llm-agents 4 llm-evaluation 1 logic-programming 1 long-context 80 looped-transformer 10 looped-transformers 1 LoRA 3 low-rank 8 masked-diffusion 1 mechanism-design 1 mechanistic-interpretability 6 memorization 3 memory 10 memory-efficiency 1 meta-learning 3 mixed-precision 1 mixture-of-experts 2 model-compression 2 model-conversion 1 moe 1 multi-agent 3 multilingual 2 multimodal 1 neural-networks 1 neurosymbolic 1 numerical-stability 1 online-learning 1 online-optimization 2 optimal-transport 1 optimization 2 optimizers 1 option-caching 1 parallel-training 1 parametric-memory 2 parity 1 persistent-state 1 phase-aware-attention 1 position-encoding 1 positional-encoding 5 post-training 1 preconditioning 1 prefix-caching 1 pretraining 10 privacy 1 prompt-order 1 prompt-repetition 2 quantization 10 random-features 1 reasoning 9 recurrence 1 recurrent-depth 1 recurrent-memory 46 recurrent-model 1 recurrent-state 1 recursive-self-improvement 1 reinforcement-learning 8 reparameterization 1 representation-geometry 1 representation-learning 2 representation-similarity 1 residual-connections 1 residual-learning 1 residual-stream 2 retrieval 15 rnn 1 robustness 1 rope 1 routing 1 sampling 2 scaling-laws 2 score-matching 1 selective-reading 1 self-conditioning 2 self-improvement 1 self-modification 1 sequence-modeling 2 set-functions 1 small-language-models 1 softmax 1 sorting 1 sparse-attention 19 sparse-autoencoders 1 sparse-inference 1 sparse-memory 2 sparsity 1 spectral-analysis 2 spectral-optimization 1 speculative-decoding 1 stability 2 state-space-models 7 state-tracking 11 structured-layers 1 structured-pruning 1 tabular-generation 1 tensor-algebra 1 test-time-regression 2 test-time-training 6 text-generation 2 theorem-proving 1 token-eviction 1 token-identity 1 token-selection 1 tokenization 3 training-efficiency 1 transformer 31 transformer-learning 1 unlearning 1 value-residual 1 vector-quantization 2 vector-search 1 vision 2 vision-language-models 1 vocabulary-expansion 1 working-memory 1