AI Research

Autoencoders vs. PCA: I Rigged the Test and PCA Still Won

A theoretical advantage that didn't survive contact with a real benchmark. The post Autoencoders vs. PCA: I Rigged the Test and PCA Still Won appeared first on Towards Data Science.

Can an Apartment Search Agent Call the Model Fewer Times and Still Find Good Matches?

I traced 2,500 listing checks with Weights & Biases Weave, removed avoidable model work one change at a time, and scored every version against the same answers. The post Can an Apartment Search Agent Call the Model…

What the ReLU Revolution Revealed About Biological Plausibility

The activation function was never a fixed biological commitment but a working hypothesis revised under empirical pressure. The post What the ReLU Revolution Revealed About Biological Plausibility appeared first on…

AI in Science, Math: A Monthly Guide to AI-Enabled Progress — September 2026

From genome research and laboratory-tested molecules to mathematical proofs and open scientific datasets, September brought a wide range of AI-enabled progress. This monthly guide covers 56 developments and explains…

a16z State of Markets 2026: AI Became a Capital Cycle

a16z’s 90-page State of Markets report argues AI has escaped the software industry and become a macroeconomic force, reshaping capital spending, debt, labor, power, hardware, and markets.

Your AI Bill Is a Toll Booth. Stop Paying Twice.

The budget nobody saw coming The post Your AI Bill Is a Toll Booth. Stop Paying Twice. appeared first on Towards Data Science.

China Is Regulating How AI Companions Build Emotional Attachment

China is setting rules for how AI companions develop emotional relationships and respond when users try to leave. The requirements address manipulation and risks to minors, while distress monitoring raises questions…

😸 Trump renamed AI "Super Intelligence"

Engineering Efficient Self-Play Chess: Search, Replay, and Throughput Under Limited Compute

arXiv:2609.37447v1 Announce Type: new Abstract: How strong can an AlphaZero-style chess system become under limited training compute when its entire learning loop is engineered for efficiency? We train from random…

Width Expansion as a Method for Class Incremental Learning

arXiv:2609.37702v1 Announce Type: new Abstract: Class Incremental Learning (Class-IL) requires models to learn new classes over time while preserving previously acquired knowledge without access to past data or task…

HyDI: A hybrid Deep Learning-Inductive Logic Programming ensemble for multi-label classification

arXiv:2609.37740v1 Announce Type: new Abstract: While attaining remarkable results for many applications, Deep Learning models are notoriously difficult to explain. This work introduces HyDI, a hybrid ensemble…

Scale Sensitivity in Low-Bit Post-Training Quantization: Curvature of the Quantization Error Landscape

arXiv:2609.37416v1 Announce Type: new Abstract: Post-training quantization (PTQ) methods in the GPTQ family minimize a layer-wise reconstruction error on a uniform grid whose scale must be chosen; the common max-based…

Learning Macroscopic Dynamics without Reconstructing Microscopic States

arXiv:2609.37392v1 Announce Type: new Abstract: Modeling the temporal evolution of macroscopic properties of complex systems is an important scientific task. To predict this evolution without full microscopic…

CipherGenome: Homomorphic Inference for Genomic Mixture-of-Experts

arXiv:2609.35883v1 Announce Type: cross Abstract: Genome foundation models are growing into sparse mixture-of-experts (MoE) networks whose expert weights no longer fit on the machines that hold the sequences, yet…

From Checkpoint Variation to Selection Gains in Supervised Fine-Tuning

arXiv:2609.36569v1 Announce Type: new Abstract: Checkpoint selection is a routine decision in supervised fine-tuning (SFT): training produces multiple checkpoints, but only one is retained. Yet fixed-budget comparisons…

Optimal Quantum-Classical Separations for Exact Learning

arXiv:2609.38073v1 Announce Type: cross Abstract: We study exact learning with membership queries for concept classes $\mathcal C\subseteq\{0,1\}^N$, focusing on the relationships among their deterministic, randomized,…

Fooling Algorithms in Non-Stationary Bandits using Belief Inertia

arXiv:2511.05620v2 Announce Type: replace Abstract: We study worst-case dynamic regret of specific multi-armed bandit algorithms on piecewise-stationary instances with at most one breakpoint. Our constructions exploit…

Privy to the Foil: Recasting Value Estimation with a Self-Privileged Critic for RLVR

arXiv:2609.37825v1 Announce Type: new Abstract: Assigning credit to intermediate steps remains a central challenge in training Large Language Models (LLMs) on multi-step reasoning tasks with sparse terminal rewards, and…

Making Cross-Continental Federated Learning Repeatable with FLIP: a Multi-Application Study

arXiv:2609.36001v1 Announce Type: new Abstract: Federated learning (FL) in healthcare remains challenging, as the overhead of rebuilding governance guarantees for every collaboration stops most projects at the…

Delta-Matching: Closing the Final Gap of Native 8-bit Training for LLMs

arXiv:2609.37852v1 Announce Type: new Abstract: Reliable FP8 attention remains a barrier to fully native 8-bit large language model training. We derive how forward-backward inconsistencies produce stale delta and…