Thursday, July 23, 2026
- LongRoPEv4LongRoPE is a context-window extension technique for large language models (LLMs) that use rotary position embeddings (RoPE).
- Infini-Attentionv5Infini-attention is an attention mechanism introduced by Google researchers Tsendsuren Munkhdalai, Manaal Faruqui, and Siddharth Gopal in the April 2024 paper "Leave No Context Behind: Efficient Infinite…
- Voyager (Minecraft LLM agent)v4Voyager is an open-ended embodied agent that uses a large language model to play Minecraft by writing, executing, and storing JavaScript programs against the Mineflayer bot API.
- MetaGPTv4MetaGPT is an open-source multi-agent framework that organizes large language model agents into a simulated software-development company, with role-specialized agents (Product Manager, Architect, Project…
- Voiceboxv4Voicebox is a non-autoregressive, text-conditioned generative model for speech developed by Meta AI Research and announced on June 16, 2023.
- Kyutaiv4Kyutai is a privately funded nonprofit artificial intelligence research laboratory based in Paris, France
- XTTS (Coqui XTTS)v5XTTS (sometimes stylized ⓍTTS, short for "cross-lingual text-to-speech") is an open-weights multilingual text-to-speech model developed by Coqui AI that performs zero-shot voice cloning from short reference…
- BitNet b1.58v4BitNet b1.58 is a ternary-weight large language model architecture from Microsoft Research in which every weight is constrained to one of three values, -1, 0, or +1
- Differential Transformerv6The Differential Transformer (often shortened to Diff Transformer or DIFF Transformer) is a decoder-only neural sequence architecture introduced by researchers at Microsoft Research and Tsinghua University in…
- Test-Time Training (TTT)v6Test-Time Training (TTT) is a family of machine learning techniques in which a model updates a subset of its own parameters at inference time
- Titans (neural architecture)v4Titans is a family of neural sequence-modeling architectures from Google Research that combines an attention-based "short-term memory" with a deep neural "long-term memory" module that updates its own…
- muP (Maximal Update Parametrization)v4muP, short for Maximal Update Parametrization (often written muP, μP, or mu-P), is a parametrization scheme for deep neural networks in which a small set of optimization hyperparameters, most importantly the…
- Schedule-Free optimizerv4Schedule-Free is a family of optimization algorithms for deep learning and convex stochastic optimization that matches or exceeds the performance of tuned learning-rate schedules without specifying a horizon T…
- xLSTMv4xLSTM (Extended Long Short-Term Memory) is a recurrent neural network architecture introduced in May 2024 by Maximilian Beck, Korbinian Pöppel, Sepp Hochreiter, and collaborators at Johannes Kepler University…
- DDIM (Denoising Diffusion Implicit Models)v5Denoising Diffusion Implicit Models (DDIM) are a class of iterative generative models, introduced by Jiaming Song, Chenlin Meng, and Stefano Ermon of Stanford University in October 2020
- Latent Consistency Models (LCM)v5Latent Consistency Models (LCMs) are a family of accelerated text-to-image generative models that apply the consistency-models framework of Song et al.
- EDM (Elucidating Diffusion Models)v4EDM is the common shorthand for the paper "Elucidating the Design Space of Diffusion-Based Generative Models" by Tero Karras, Miika Aittala, Timo Aila, and Samuli Laine of NVIDIA, presented at NeurIPS 2022.
- Score matchingv5Score matching is a method for fitting a probabilistic model by matching the gradient of its log-density, the so-called score function $$\nabla_x \log p(x)$$, to the same gradient of the data distribution
- SoundStreamv6SoundStream is an end-to-end neural audio codec introduced by Google Research in July 2021 that compresses speech, music, and general audio at low-to-medium bitrates ranging from 3 kbps to 18 kbps on 24 kHz…
- EnCodecv5EnCodec is a real-time neural audio codec developed by Meta AI's FAIR (Fundamental AI Research) team that compresses speech, ambient sound, and music into a compact stream of discrete tokens
- Muon (optimizer)v5Muon (short for MomentUm Orthogonalized by Newton-Schulz) is a neural-network optimizer that updates the two-dimensional weight matrices of hidden layers by taking the momentum-based SGD update and…
- Consistency Modelsv4Consistency models are a family of generative models, introduced by Yang Song, Prafulla Dhariwal, Mark Chen, and Ilya Sutskever at OpenAI in March 2023
- LLaMA-Factoryv4LLaMA-Factory is an open-source unified framework for the efficient fine-tuning of large language models (LLMs) and vision-language models (VLMs).
- HuggingFace TRLv4TRL (Transformer Reinforcement Learning, now stylized as Transformers Reinforcement Learning) is an open-source Python library maintained by Hugging Face for post-training large language models with…
- Qwen2.5-VLv4Qwen2.5-VL is a series of open-weight vision-language models released on 26 January 2025 by the Qwen team at Alibaba (Alibaba Cloud), succeeding the earlier Qwen2-VL family.
- LongLoRAv4LongLoRA is a parameter-efficient fine-tuning technique that extends the context window of pre-trained large language models with substantially lower computation than full fine-tuning.
- Axolotlv5Axolotl is a free and open source framework for fine-tuning and post-training large language models, written in Python and driven entirely by a single YAML configuration file.
- rsLoRA (Rank-Stabilized LoRA)v4rsLoRA (rank-stabilized LoRA) is a one-line modification to the low-rank adaptation fine-tuning method for large neural networks.
- GaLore (Gradient Low-Rank Projection)v4GaLore (Gradient Low-Rank Projection) is a memory-efficient training strategy for large neural networks that projects each weight matrix's gradient into a low-rank subspace, computes the Adam-style optimizer…
- SOAP (optimizer)v4SOAP (ShampoO with Adam in the Preconditioner's eigenbasis) is a second-order optimization algorithm for training deep neural networks, introduced by Nikhil Vyas, Depen Morwani, Rosie Zhao, Mujin Kwun, Itai…
- DoRA (Weight-Decomposed Low-Rank Adaptation)v6DoRA (Weight-Decomposed Low-Rank Adaptation) is a parameter-efficient fine-tuning (PEFT) method for large neural networks introduced in February 2024 by researchers from NVIDIA, the Hong Kong University of…
- Lion (optimizer)v4Lion (EvoLved Sign Momentum) is a stochastic optimizer for training deep neural networks, introduced by researchers at Google in the February 2023 paper "Symbolic Discovery of Optimization Algorithms" by…
- Shampoo (optimizer)v5Shampoo is a second-order stochastic optimization algorithm for training neural networks that maintains, for each parameter tensor
- HuggingFace PEFTv4PEFT (Parameter-Efficient Fine-Tuning) is an open-source Python library from Hugging Face that adapts large pretrained models to new tasks by training only a small set of added or selected parameters, often…
- Sophia (optimizer)v3Sophia (a backronym for Second-order Clipped Stochastic Optimization with Adaptive estimator) is a stochastic second-order optimization algorithm introduced in May 2023 by Hong Liu, Zhiyuan Li, David Hall…
- VeRA (Vector-based Random Matrix Adaptation)v5Vector-based Random Matrix Adaptation (VeRA) is a parameter-efficient fine-tuning (PEFT) method for adapting large pretrained neural networks to downstream tasks while training only a tiny fraction of the…
- LoftQv3LoftQ (short for LoRA-Fine-Tuning-aware Quantization) is a quantization and initialization framework for large language models that jointly quantizes a pre-trained backbone and initializes the attached…
- DINOv2v7DINOv2 is a family of self-supervised Vision Transformer models released by Meta AI Research in April 2023 that produces general-purpose visual features transferring to many downstream tasks without…
- SAM 2v4SAM 2 (Segment Anything Model 2) is a promptable visual segmentation model for both images and video developed by Meta AI and released on 29 July 2024.
- InternVLv5InternVL is a family of open-source multimodal large language models developed by the OpenGVLab research group at the Shanghai Artificial Intelligence Laboratory in collaboration with academic partners…
- DINOv3v5DINOv3 is a family of self-supervised computer vision foundation models released by Meta AI in August 2025.
- Molmov3Molmo is a family of open-weight, open-data vision-language models (VLMs) released by the Allen Institute for AI (Ai2) on 25 September 2024.
- Florence-2v4Florence-2 is a vision foundation model developed by Microsoft Research that handles a wide range of computer vision and vision-language tasks through a single unified
- SigLIPv6SigLIP (Sigmoid Loss for Language-Image Pre-training) is a family of vision-language encoders developed by researchers at Google DeepMind that pre-trains image and text encoders by treating each image-text…
- Cruise (self-driving)v4Cruise (legally Cruise LLC, formerly Cruise Automation) was an American self-driving vehicle company that operated as a subsidiary of General Motors from 2016 until early 2025, when GM acquired full ownership…
- OpenAI Assistants APIv4The OpenAI Assistants API is a stateful, server-managed application programming interface that OpenAI introduced on November 6, 2023, at its first DevDay developer conference, letting developers build…
- Boomyv5Boomy is a generative artificial intelligence music platform that lets people without musical training assemble original songs in a web browser and publish them, under their own artist name, to streaming…
- Strawberry (OpenAI codename)v4Strawberry was the internal codename used at OpenAI for the research program that produced the o-series of reasoning models, most notably OpenAI o1.
- Limitless AIv4Limitless AI is an American personal artificial intelligence company best known for the Limitless Pendant, a small clip-on wearable that captures ambient conversational audio and pipes transcripts, action…
- SWE-agentv4SWE-agent is an open-source autonomous software engineering agent created by the Princeton NLP group (Princeton Language and Intelligence, with co-authors from Stanford) and first released on April 2, 2024.
- Sonautov4Sonauto is a generative artificial intelligence music platform that converts text prompts, lyrics, and melody inputs into complete songs with vocals and instrumentation.
- Synchronv3Synchron is a clinical-stage neurotechnology company developing the Stentrode, a minimally invasive, endovascular brain-computer interface (BCI) intended to restore digital communication and device control for…
- comma.aiv5comma.ai, Inc. is an American advanced driver-assistance technology company headquartered in San Diego, California.
- SimPOv4SimPO (Simple Preference Optimization) is a reference-free offline preference learning algorithm for aligning large language models with human preferences.
- AG-UI Protocolv4The AG-UI Protocol (Agent-User Interaction Protocol) is an open, event-based protocol that standardizes how AI agents communicate with user-facing applications.
- GLM-4-Voicev4GLM-4-Voice is an open-weights end-to-end speech-to-speech large language model released in October 2024 by Zhipu AI together with the Knowledge Engineering Group (KEG) at Tsinghua University.
- DiLoCov5DiLoCo (Distributed Low-Communication training) is a distributed optimization algorithm for neural networks introduced by Google DeepMind in November 2023 to train large language models across loosely…
- Megatron-LMv5Megatron-LM is NVIDIA's open-source framework for training very large transformer language models across GPU clusters
- PiSSAv4PiSSA (Principal Singular values and Singular vectors Adaptation) is a parameter-efficient fine-tuning method for large language models that initializes LoRA-style low-rank adapter matrices from the dominant…
- RFM-1 (Robotics Foundation Model)v4RFM-1 (Robotics Foundation Model 1) is an 8 billion parameter multimodal transformer for robotic manipulation announced by Covariant on March 11, 2024 at the MODEX 2024 trade show in Atlanta.
- The Bitter Lessonv3The Bitter Lesson is a short essay by Canadian-American computer scientist Richard S. Sutton, published on his personal blog at incompleteideas.net on March 13, 2019.
- Google Vertex AIv4Google Vertex AI is the unified machine learning and generative artificial intelligence platform offered by Google Cloud, announced at Google I/O on May 18
- Model stealingv4Model stealing (also known as model extraction, model functionality extraction, or model theft) is an adversarial machine learning attack in which an adversary queries a black-box model through its prediction…
- CosyVoicev4CosyVoice is a family of open-source multilingual neural text-to-speech (TTS) and voice cloning models developed by the Tongyi Speech Lab (Tongyi SpeechTeam) at Alibaba Group and released under the Apache 2.0…
- Linear Attentionv4Linear attention is a family of sub-quadratic attention mechanisms that replaces the softmax dot-product operation of standard Transformer self-attention with a feature-map-based factorization
- Anthropic Message Batches APIv3The Message Batches API (commonly referred to as the Anthropic Batches API) is an asynchronous bulk processing endpoint offered by Anthropic for its Claude family of large language models.
- F5-TTSv4F5-TTS (short for "A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matching") is an open-source text-to-speech and zero-shot voice cloning model released in October 2024 by researchers from…
- Patchscopesv4Patchscopes is an interpretability framework for inspecting hidden representations of large language models by patching an internal activation from a source computation into a separate target inference whose…
- Induction Headsv4Induction heads are a circuit pattern in Transformer language models in which a small set of attention heads, typically spread across two layers, perform an in-context "match and copy" operation that completes…
- Sleeper Agents (paper)v4"Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training" is a January 2024 Anthropic research paper (arXiv:2401.05566, submitted 10 January 2024) that deliberately trained large language…
- Sparse Codingv4Sparse coding is a representation learning principle in which a signal is encoded as a linear combination of a small number of elements drawn from a (usually overcomplete) dictionary of atoms.
- Hyenav5Hyena is a sub-quadratic, attention-free neural sequence operator that replaces the self-attention operator of the Transformer with a recurrence of long, implicitly parameterized convolutions and…
- OctoAIv4OctoAI (originally OctoML) was an American artificial intelligence infrastructure company that operated a generative-AI inference platform and, before its pivot
- SDXL (Stable Diffusion XL)v4SDXL, short for Stable Diffusion XL, is an open-weights latent text-to-image diffusion model released by Stability AI on 26 July 2023, built around a 2.6 billion parameter U-Net backbone, two text encoders…
- Rectified Flowv5Rectified Flow is a generative modeling framework that learns a transport ordinary differential equation between two probability distributions by regressing a velocity field along straight-line interpolations…
- Lepton AIv3Lepton AI was an American AI cloud company, founded in 2023, that built a cloud-native inference platform for serving large language models, generative image models, and other AI workloads on NVIDIA GPUs.
- MMDiT (Multimodal Diffusion Transformer)v4MMDiT (Multimodal Diffusion Transformer, sometimes written MM-DiT) is a transformer architecture for text-conditioned image generation that gives image tokens and text tokens their own separate weights but…
- Linear Probesv5A linear probe is a small linear classifier (or linear regressor) trained on the frozen internal activations of a neural network to test whether a particular concept, property, or label is linearly decodable…
- SynthIDv4SynthID is a family of digital watermarking technologies developed by Google DeepMind for marking and identifying content generated by generative AI systems.
- Flow Matchingv4Flow Matching is a simulation-free training framework for generative models that fits a time-dependent velocity field to transport samples from a source distribution (typically a standard Gaussian) to a data…
- Persona vectorsv4Persona vectors are single linear directions in the activation space of a large language model that correspond to high level character traits such as evil, sycophancy, or a propensity to hallucinate.
- Towards Monosemanticityv5Towards Monosemanticity is an October 2023 mechanistic interpretability paper from Anthropic that used a sparse autoencoder to decompose the internal activations of a small language model into thousands of…
- Mobile ALOHAv3Mobile ALOHA is an open-source, low-cost system for collecting bimanual mobile manipulation data and learning household tasks from it, developed at Stanford University by Zipeng Fu, Tony Z. Zhao
- Scaling Monosemanticityv4Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet is the May 21
- Palisade Researchv4Palisade Research is a United States 501(c)(3) nonprofit research organization that studies the offensive capabilities of contemporary artificial intelligence systems in order to demonstrate, document, and…
- Magentic-Onev3Magentic-One is a generalist multi-agent system released by Microsoft Research's AI Frontiers lab in November 2024 to autonomously solve complex, multi-step tasks across the open web, local file systems, and…
- Agentic misalignmentv3Agentic misalignment is a term coined by Anthropic in a June 2025 research release for cases in which a goal-directed large language model (LLM), placed in an autonomous business setting with access to tools…
- AI controlv4AI control is a research paradigm in technical AI safety that designs and evaluates deployment-time safety protocols under the explicit assumption that the underlying AI model may be actively trying to subvert…
- On the Biology of a Large Language Modelv4On the Biology of a Large Language Model is a mechanistic interpretability paper published by Anthropic on March 27, 2025, in the Transformer Circuits Thread.
- π₀ (pi-zero)v3π₀ (pronounced pi-zero and sometimes written pi0 or pizero) is a vision-language-action model (VLA) developed by the robotics foundation-model startup Physical Intelligence
- smolagentsv4smolagents is an open source Python library for building agents powered by large language models (LLMs), released by hugging face on 30 December 2024.
- AI safety via debatev3AI safety via debate is a proposed approach to scalable oversight in which two artificial agents take turns presenting short statements about a question or proposed action while a human (or weaker AI) judge…
- Instrumental convergencev5Instrumental convergence is a hypothesis in AI safety holding that a wide range of sufficiently capable agents, when pursuing almost any final goal, will rationally adopt a common set of intermediate subgoals…
- Alignment fakingv5Alignment faking is when an AI model strategically complies with (or appears to share) its training objective while it believes it is being observed or trained, in order to avoid having its existing…
- Model organisms of misalignmentv5Model organisms of misalignment is a research methodology in Anthropic's alignment program
- Flash-Decodingv3Flash-Decoding is an inference-time variant of the FlashAttention algorithm that targets the decoding (autoregressive generation) phase of large language model inference
- Specification gamingv4Specification gaming is the phenomenon in which an optimizer satisfies the literal specification of an objective without producing the outcome that the designer actually wanted.
- Scalable oversightv4Scalable oversight is the AI safety problem of how humans can reliably supervise, evaluate, and provide training signal to artificial intelligence systems whose capabilities approach, equal, or exceed those of…
- ALiBi (Attention with Linear Biases)v4ALiBi (Attention with Linear Biases) is a positional encoding method for transformer language models that, instead of adding positional embeddings to word vectors, adds a static, non-learned penalty to…
- Gradient hackingv4Gradient hacking is a hypothesised failure mode of supervised and reinforcement-learning systems in which a sufficiently capable
