Notable data and AI papers

Papers627Labs45

627 papers · 45 labs

Physics of Language Models: Part 4.1, Architecture Design and the Magic of Canon Layers research paper by Meta, 2025

Meta · Architectures · Dec 19, 2025 · 48 citations · 29 upvotes · unverified

Read on arXiv

What it shows

Controlled synthetic pretraining reveals CANON LAYERS, lightweight architectural components that enhance information flow and model capabilities across various sequence architectures.

By Zeyuan Allen-Zhu · arXiv 2512.17351 · PDF · Code

UnverifiedHugging Face's summary; not yet checked by hand.

More from Meta

All 40
Multimodal and roboticsUnverifiedOct 25, 2026InteractiveWorld: A Verifiable Human-Centric Environment for Building Better Interactive MLLM AssistantsInteractiveWorld provides a verifiable simulation environment and benchmark to evaluate whether multimodal language models can reliably guide humans through physical tasks.Meta46 upvotesRead →Multimodal and roboticsUnverifiedOct 2, 2026Native Action-Prior Learning from Videos for World Action ModelsNAVA-WAM learns robot action priors directly from observation-only videos, improving control performance without needing action-labeled data for pretraining.Meta88 upvotesRead →Alignment and safetyUnverifiedOct 1, 2026Sharpening Tax in Post-TrainingThis paper introduces a metric to quantify the loss in solution coverage after reinforcement learning and proposes a sampler to mitigate it.Meta1 citations111 upvotesRead →Multimodal and roboticsUnverifiedOct 1, 2026World Action Modeling with Progressive Visual PlanningProWAM improves robotic control by predicting sparse visual sub-goals to guide action generation, avoiding the need for dense video generation.Meta1 citations98 upvotesRead →ArchitecturesUnverifiedSep 29, 2026Context Language ModelsContext Language Models manage their own context by treating it as an editable file, improving performance and efficiency.Meta5 citations47 upvotesRead →Inference and efficiencyUnverifiedSep 26, 2026In-Flight KV Cache with Clean Anchors for Faster Autoregressive Video DiffusionFlashForward accelerates video diffusion by reusing in-flight KV caches and using sparse clean anchors to maintain structural consistency.Meta1 citations34 upvotesRead →Agents and evaluationUnverifiedSep 4, 2026WearableQA: A Benchmark for Health Reasoning over Real-World Wearable DataWearableQA is a benchmark of multiple-choice questions derived from real longitudinal wearable data that evaluates large language model reasoning across data and health dimensions.Meta43 upvotesRead →Multimodal and roboticsAug 6, 2026Towards Physics of Multimodal Pretraining: Knowledge Flow, Modality Synergy, Early Unification, and RecipesControlled experiments on when text, image understanding and image generation help or compete when trained together.Meta7 citations63 upvotesRead →

Other architectures papers

Topic
ArchitecturesJun 12, 2017Attention Is All You NeedIntroduced the Transformer, the attention-only architecture behind nearly every large language model since.Google201k citations142 upvotesRead →ArchitecturesDec 1, 2023Mamba: Linear-Time Sequence Modeling with Selective State SpacesA selective state space model that scales linearly with sequence length and matches Transformers on language.Carnegie Mellon University10.2k citations152 upvotesRead →ArchitecturesSep 2, 2021CodeT5: Identifier-aware Unified Pre-trained Encoder-Decoder Models for Code Understanding and GenerationCodeT5, a unified encoder-decoder Transformer model, improves code understanding and generation by leveraging semantic information from identifiers and user comments.Salesforce2,633 citations3 upvotesRead →ArchitecturesUnverifiedJan 11, 2024DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language ModelsThe DeepSeekMoE architecture improves expert specialization in Mixture-of-Experts models by segmenting experts and isolating shared ones, achieving better performance and computational efficiency compared to GShard and other models.DeepSeek1,268 citations65 upvotesRead →ArchitecturesMay 13, 2023CodeT5+: Open Code Large Language Models for Code Understanding and GenerationCodeT5+, a family of flexible encoder-decoder LLMs initialized with frozen LLMs and trained with a mixture of pretraining objectives, achieves state-of-the-art performance across various code-related tasks.Salesforce836 citations7 upvotesRead →ArchitecturesUnverifiedOct 9, 2019HuggingFace's Transformers: State-of-the-art Natural Language ProcessingTransformers library provides state-of-the-art Transformer architectures and pretrained models for natural language processing tasks with a unified API and emphasis on extensibility and robust deployment.Hugging Face251 citations34 upvotesRead →
About this paper
Venue
Neural Information Processing Systems
Citations
48, 3 influential · Semantic Scholar
Upvotes
29 · Hugging Face
Lab
Meta · on Companies · on Acquisitions · on Quarterly · on Paydays · on TechConf

Changes

CitationsNov 9, 2026Citations rose from 47 to 48.
CitationsNov 2, 2026Citations rose from 44 to 47.
CitationsOct 5, 2026Citations rose from 41 to 44.
CitationsSep 28, 2026Citations rose from 40 to 41.
CitationsSep 25, 2026A first count of 3 influential citations.
CitationsSep 25, 2026A first count of 40 citations.
New paperSep 25, 2026Found by the weekly Hugging Face scan; not yet checked by hand.