Entity Pages

Files

• act-2023.md - Action Chunking with Transformers method for continuous robot action chunks.
• anomod-2026.md - Multimodal microservice anomaly-detection and root-cause-analysis dataset.
• armt-2024.md - Associative Recurrent Memory Transformer method extending RMT with layerwise associative memory.
• atlas-2025.md - ATLAS test-time memory module and DeepTransformers family.
• boom-2025.md - Datadog observability metrics forecasting benchmark.
• bolmo-2025.md - Fully open byte-level LM family produced by byteifying subword LMs.
• bittokens-2025.md - Bit-level single-token number encoding method based on IEEE 754 features.
• charm-2025.md - Channel-aware JEPA embedding model for multivariate time series.
• chatts-2024.md - Time-series multimodal LLM trained from synthetic time-series/text data.
• chronograph-2025.md - Graph-native multivariate microservice time-series benchmark with temporal node and edge features.
• compute-optimal-tokenization-2026.md - Tokenization-aware scaling-law study centered on bytes per parameter and compute-optimal compression rate.
• context-is-key-2024.md - ServiceNow benchmark for probabilistic forecasting with essential natural-language context.
• cwm-2025.md - Meta FAIR Code World Model for code generation through computational-environment action-observation trajectories.
• diffusionblocks-2026.md - Block-wise training framework that reinterprets residual-network depth as diffusion-style denoising blocks.
• diffusion-policy-2023.md - Visuomotor diffusion policy over future continuous action trajectories.
• dinov3-2025.md - Self-supervised vision foundation model suite.
• dragon-hatchling-2025.md - Pathway BDH / Dragon Hatchling recurrent attention/state-space architecture.
• ebt-2025.md - Energy-Based Transformer method for candidate-prediction scoring, gradient-based refinement, and dynamic compute.
• eggroll-2025.md - Low-rank perturbation method for hyperscale evolution strategies.
• fade-2026.md - Adaptive per-parameter weight-decay method for controlled forgetting in continual learning.
• eidos-2026.md - Time-series foundation model family based on latent-space predictive learning.
• elt-2026.md - Elastic Looped Transformer architecture for any-time visual generation.
• flowstate-2025.md - SSM-based time-series foundation model with continuous functional-basis decoding.
• florence-2-2023.md - Microsoft prompt-based vision foundation model built around the FLD-5B iterative data engine.
• fone-2025.md - Fourier Number Embedding method for single-token number representations.
• gaia-micross-2021.md - GAIA AIOps collection and MicroSS microservice telemetry subset.
• fast-2025.md - Frequency-space tokenizer for robot action chunks.
• gemma-4-12b-2026.md - Google DeepMind encoder-free multimodal 12B open-weight model.
• gemini-robotics-1-5-2025.md - Google DeepMind embodied-reasoning and VLA robot model family.
• genie-2024.md - Google DeepMind generative interactive environment model with learned latent actions from unlabeled video.
• gift-eval-2024.md - General-purpose time-series forecasting benchmark and leaderboard.
• gr00t-n1-2025.md - NVIDIA open humanoid VLA/action model.
• h-net-2025.md - Hierarchical sequence model with learned dynamic chunking.
• helix-2025.md - Figure AI upper-body humanoid VLA with fast/slow control.
• helix-02-2026.md - Figure AI full-body humanoid VLA/controller stack.
• hierarchical-reasoning-model-2025.md - HRM fast/slow recurrent reasoning architecture.
• hyperloop-transformers-2026.md - Looped language-model architecture using loop-level hyper-connections.
• huginn-2025.md - Recurrent-depth language model using latent-space loops as test-time compute.
• latent-thoughts-2025.md - Looped Transformer framing for latent thought steps.
• lemma-rca-2024.md - Large multi-modal multi-domain root-cause-analysis dataset collection.
• llm-sleep-2026.md - Sleep-time memory-consolidation method for SSM-attention hybrid language models.
• lejepa-2025.md - JEPA objective combining predictive alignment with SIGReg.
• leautoencoder-2026.md - Self-teaching autoencoder prototype using transformed latent consistency instead of direct image-space reconstruction loss.
• loopformer-2026.md - Elastic-depth looped Transformer for budget-conditioned latent reasoning.
• mamba-2023.md - Selective state space model architecture for efficient recurrent sequence modeling.
• mamba-2-2024.md - Structured state space duality architecture and SSD algorithm.
• mamba-3-2026.md - Mamba-family architecture with richer discretization, complex state, and MIMO inference updates.
• mantis-2025.md - Time-series classification foundation-model lineage covering Mantis, MantisV2, and UTICA.
• mesanet-2025.md - Sequence model with locally optimal test-time training.
• mhc-2025.md - Manifold-Constrained Hyper-Connections method for stable matrix-valued residual streams.
• miras-2025.md - Associative-memory framework for attentional bias, retention, and online optimization.
• moda-2026.md - Mixture-of-Depths Attention method for content-based inter-layer depth retrieval.
• moirai-2024.md - Salesforce Uni2TS forecasting family covering Moirai 1.x, Moirai-MoE, and Moirai 2.0.
• octo-2024.md - Open-source generalist robot policy.
• openvla-2024.md - Open 7B VLA model using discretized action tokens.
• openrca-2025.md - LLM-agent root-cause-analysis benchmark over large software telemetry.
• ops-lite-2026.md - Compact RCA evaluation set with per-case causal graph ground truth.
• pdr-rtv-2026.md - Agentic-coding test-time scaling recipe using structured summaries, recursive tournament voting, and refinement.
• parallel-samplers-recurrent-depth-2025.md - Parallel inference method for recurrent-depth models.
• pararnn-2025.md - Parallel nonlinear RNN training framework from Apple.
• parcae-2026.md - Stable looped language-model architecture with scaling-law analysis.
• perception-encoder-2025.md - Meta vision-encoder family whose strongest general features are often internal before alignment tuning.
• pi0-2024.md - Physical Intelligence VLA flow model for general robot control.
• pi0-7-2026.md - Steerable generalist VLA model with metadata, subgoal images, and flow action expert.
• raev2-2026.md - Multi-layer representation-autoencoder recipe for generation and navigation world-model rollouts.
• rcaeval-2025.md - Microservice root-cause-analysis benchmark and evaluation framework.
• rdt-1b-2024.md - Robotics Diffusion Transformer for bimanual manipulation.
• rate-2023.md - Recurrent Action Transformer with Memory offline RL policy architecture.
• recurrent-transformer-2026.md - Transformer variant with layerwise recurrent memory.
• rmt-2022.md - Recurrent Memory Transformer segment-level memory-token method.
• reinpatch-2026.md - Learned adaptive patching method for time-series forecasting.
• rt-2-2023.md - Vision-language-action model using action-as-language tokens.
• rwkv-ts-2024.md - RWKV-style recurrent sequence model adapted to time-series tasks.
• simmtm-2023.md - Masked time-series modeling framework based on multi-neighbor reconstruction.
• sparse-layers-looped-language-models-2026.md - Sparse MoE looped language-model branch.
• stable-worldmodel-2026.md - Reproducible world-model research platform with trajectory data handling, planning solvers, baselines, and factor-of-variation evaluation.
• sundial-2025.md - THUML flow-matching time-series forecasting foundation-model family.
• t2s-2025.md - Text-to-time-series generation model with LA-VAE and flow-matching Diffusion Transformer.
• tabm-2024.md - MLP-based tabular deep-learning model with parameter-efficient ensembling.
• telecomts-2025.md - Multimodal 5G observability benchmark for anomaly detection, root-cause analysis, forecasting, and time-series/text Q&A.
• time-2026.md - Contamination-resistant zero-shot forecasting benchmark.
• time-hd-2025.md - High-dimensional time-series forecasting benchmark introduced with U-Cast.
• time-series-library-2024.md - Time-series benchmark collection and LSF/LTSF dataset handle.
• timeomni-1-2026.md - Time-series reasoning model and TSR-Suite.
• timeomni-vl-2026.md - Vision-centric time-series understanding/generation framework.
• timesfm-2023.md - Google decoder-only time-series forecasting foundation model.
• tiny-recursive-model-2025.md - TRM tiny recursive reasoning model.
• titans-2025.md - Neural long-term memory sequence-model family.
• titans-revisited-2025.md - Titans reimplementation and critical-analysis object.
• tiny-time-mixers-2024.md - IBM Granite compact pretrained mixer-style forecasting model family.
• toto-2025.md - Datadog observability-oriented forecasting family covering Toto 1.0 and Toto 2.0.
• ts2vec-2021.md - Hierarchical contrastive time-series representation model.
• tsmixer-2023.md - All-MLP time-series forecasting architecture.
• turboquant-2025.md - Online vector quantization method for KV-cache and vector-search state, with vLLM caveats around FP8, latency, throughput, and memory pressure.
• tuna-2-2026.md - Pixel-space unified multimodal model.
• units-2024.md - Unified multi-task time-series model.
• universal-reasoning-model-2025.md - UT-derived recursive reasoning model.
• universal-transformers-2018.md - Root recurrent-depth self-attention model.
• universal-transformers-need-memory-2026.md - Study of UT memory tokens and adaptive-depth tradeoffs.
• vl-jepa-2025.md - Vision-language JEPA system.