AI Wiki — Knowledge Base

全网都在吹的LeCun新论文，90%的解读都是错的。他们说生成式AI是死路，说过去三年花的几百亿全白费了，说15M参数的小模型就能吊打万亿大模型。这些全是营销号的夸张，我觉得这篇论文的真正分量比他们吹的还要重。 Yann LeCun团队这次解决了JEPA困扰了好几年的表征坍…

全网都在吹的LeCun新论文，90%的解读都是错的。他们说生成式AI是死路，说过去三年花的几百亿全白费了，说15M参数的小模型就能吊打万亿大模型。这些全是营销号的夸张，我觉得这篇论文的真正分量比他们吹的还要重。 Yann LeCun团队这次解决了JEPA困扰了好几年的表征坍缩问题。以前的世界模型，学着学着就会把狗车人都压成一模一样的向量，什么都学不到。这次他们只加了一个极其优雅的数学正则

researchyann-lecun--x

21 Apr 2026

Search: “research”

Get ready for hotter, muggier, stormier summers

GoCoMA: Hyperbolic Multimodal Representation Fusion for Large Language Model-Generated Code Attribution

Graph Neural Networks for Graphs with Heterophily: A Survey

HABIT: Chrono-Synergia Robust Progressive Learning Framework for Composed Image Retrieval

Hard to Be Heard: Phoneme-Level ASR Analysis of Phonologically Complex, Low-Resource Endangered Languages

HeteroCache: A Dynamic Retrieval Approach to Heterogeneous KV Cache Compression for Long-Context LLM Inference

HiPrune: Hierarchical Attention for Efficient Token Pruning in Vision-Language Models

HiRAS: A Hierarchical Multi-Agent Framework for Paper-to-Code Generation and Execution

Horospherical Depth and Busemann Median on Hadamard Manifolds

How Much Cache Does Reasoning Need? Depth-Cache Tradeoffs in KV-Compressed Transformers

How Much Data is Enough? The Zeta Law of Discoverability in Biomedical Data, featuring the enigmatic Riemann zeta function

HPLT 3.0: Very Large-Scale Multilingual Resources for LLMs and MT. Mono- and Bi-lingual Data, Multilingual Evaluation, and Pre-Trained Models

HyKey: Hyperspectral Keypoint Detection and Matching in Minimally Invasive Surgery

Hyperspectral Unmixing Hierarchies

I tested @huggingface ml-intern, given the prompt 'Fine-tune a Segment Anything Model (SAM) on a useful medical dataset. Train the model, an…

ICLAD: In-Context Learning with Comparison-Guidance for Audio Deepfake Detection

I’m hearing there’s renewed lobbying in DC and in state legislatures to ban or severely restrict open-source. Like a few years ago, we’ll ne…

IMA-MoE: An Interpretable Modality-Aware Mixture-of-Experts Framework for Characterizing the Neurobiological Signatures of Binge Eating Disorder

ImpRIF: Stronger Implicit Reasoning Leads to Better Complex Instruction Following

Improving Radio Interferometry Imaging by Explicitly Modeling Cross-Domain Consistency in Reconstruction

Improving reproducibility by controlling random seed stability in machine learning based estimation via bagging

Improving Speech Recognition of Named Entities in Classroom Speech with LLM Revision and Phonetic-Semantic Context

In Search of Lost DNA Sequence Pretraining

In Situ Training of Implicit Neural Compressors for Scientific Simulations via Sketch-Based Regularization

Incentivizing Parametric Knowledge via Reinforcement Learning with Verifiable Rewards for Cross-Cultural Entity Translation

IncepDeHazeGAN: Novel Satellite Image Dehazing

Inductive Convolution Nuclear Norm Minimization for Tensor Completion with Arbitrary Sampling

Inference-Time Temporal Probability Smoothing for Stable Video Segmentation with SAM2 under Weak Prompts

Instant Colorization of Gaussian Splats

💫 Introducing NeuralSet: a simple, fast, scalable Python package for Neuro-AI 📦 pip install neuralset 📄 https://kingjr.github.io/files/ne…

Inventor recalls eye imaging breakthrough

iPhoneme: Brain-to-Text Communication for ALS Using ConformerXL Decoding

Is SAM3 ready for pathology segmentation?

Joint Distillation for Fast Likelihood Evaluation and Sampling in Flow-based Models

KaLDeX: Kalman Filter based Linear Deformable Cross Attention for Retina Vessel Segmentation

Kevin Warsh just lost me. He argues he's going to be an independent Fed Chair, but refuses to acknowledge that Trump lost the 2020 election.…

L1 Regularization Paths in Linear Models by Parametric Gaussian Message Passing

LASER: Low-Rank Activation SVD for Efficient Recursion

Latent Abstraction for Retrieval-Augmented Generation

Latent-Compressed Variational Autoencoder for Video Diffusion Models

Latent Phase-Shift Rollback: Inference-Time Error Correction via Residual Stream Monitoring and KV-Cache Steering

LBFTI: Layer-Based Facial Template Inversion for Identity-Preserving Fine-Grained Face Reconstruction

Learning residue level protein dynamics with multiscale Gaussians

Learning the Riccati solution operator for time-varying LQR via Deep Operator Networks

Learning to Correct: Calibrated Reinforcement Learning for Multi-Attempt Chain-of-Thought

Learning Unanimously Acceptable Lotteries via Queries

LEPO: nderline{L}atent Rnderline{e}asoning nderline{P}olicy nderline{O}ptimization for Large Language~Models