AI Research

How Diffusion Controller unifies and simplifies AI image generation

Algorithms & Theory

😺 OpenAI DevDay 2026 Watch Party

I Compacted 1,000 Apache Iceberg Files Into 6. Here’s What Happened to Query Performance.

Benchmarking the impact of fewer, larger files across three SQL workloads The post I Compacted 1,000 Apache Iceberg Files Into 6. Here’s What Happened to Query Performance. appeared first on Towards Data Science.

GPT-6 Astra Helped Find a Fusion Math Breakthrough. Here’s What It Proves.

Two independent papers found three families of plasma equilibria long thought unlikely to exist. One researcher credits GPT-6 Astra Pro with helping discover two of them. Here’s what the results prove and what they mean…

When All You Have Are Decoders, Every Decision Looks Like Generation

Not every decision needs a decoder, generation Is not always a decision The post When All You Have Are Decoders, Every Decision Looks Like Generation appeared first on Towards Data Science.

AI Made Data Scientists Faster. Now It’s Expanding the Job.

The real shift is bigger than productivity: AI is reshaping ownership, judgment, and the career path of data scientists. The post AI Made Data Scientists Faster. Now It’s Expanding the Job. appeared first on Towards…

AI Models Can Teach Each Other Skills Without Talking About Them?

A coding-trained AI leaves subtle traces of what it learned even when choosing ordinary words. New research shows another model can learn from those traces and emerge better at coding without ever seeing the teacher’s…

How to Design Architectural Guardrails Around AI Agents

Agent design patterns every data engineer must know The post How to Design Architectural Guardrails Around AI Agents appeared first on Towards Data Science.

Building Fair Evaluation Sets Is a Combinatorial Problem

Here’s how to solve it exactly. The post Building Fair Evaluation Sets Is a Combinatorial Problem appeared first on Towards Data Science.

😺 Meta wants to own your AI front door

Claude Sonnet 5.5 vs GPT-6 Astra: where Sonnet (surprisingly) wins

Anthropic’s newest workhorse model beats GPT-6 Astra on some agentic coding and professional-work benchmarks while charging one-fifth as much per token.

OpenAI Stopped GPT-6.1 Astra. Who Checks the Safety Gate?

OpenAI’s decision to withhold GPT-6.1 Astra shows an internal safety process can stop a model from shipping.

The AI Labs Want Oversight. They Also Control the Evidence.

Anthropic says Claude now leads 26% of its measured AI R&D work, while OpenAI has reached its “automated research intern” milestone. But neither proves AI is caught in a runaway self-improvement loop. The harder…

Hill Sampling for Test-Time Scaling: A Simple and Better Alternative to Repeated Sampling, Evolution, and Training

arXiv:2609.25510v2 Announce Type: replace Abstract: Large language models (LLMs) can improve solutions to verifiable scientific and algorithmic problems by spending additional computation at test time. Recent systems…

Beyond Empirical Support: Structured Outlier Generation via Sinkhorn Optimal Transport

arXiv:2609.31470v1 Announce Type: cross Abstract: Outliers are essential for evaluating and improving the robustness of machine learning systems, especially when future distributions may differ significantly from…

Teacher-Anchored Selection of Post-Training Quantized Models under Domain Shift

arXiv:2609.31155v1 Announce Type: new Abstract: Compressing a trained model yields a family of deployment candidates, and under domain shift the most compressed one need not be the one to deploy. We study selection over…

Nonparametric In-Context Learning under Growing Geometric Complexity: Minimax Optimality and Local Geometry-Adaptivity of Transformers

arXiv:2609.31458v1 Announce Type: cross Abstract: Transformers have become a central architecture for in-context learning (ICL), particularly through their state-of-the-art performance in large language models. This…

Proper Scoring Rules for Right-Censored Survival Data

arXiv:2606.06393v2 Announce Type: replace Abstract: Proper scoring rules provide a rigorous theoretical basis for the training and evaluation of probabilistic forecasts. In survival analysis, such forecasts describe the…

Cost-Aware Best-LLM Identification using Dueling Feedback

arXiv:2609.30360v1 Announce Type: new Abstract: Inspired by the problem of identifying the best model from a collection of large language models (LLMs) with heterogeneous querying costs, we formulate and analyse a…

Scaffold: Support Graph Theory Based Sparsification for Graph Neural Networks

arXiv:2609.31466v1 Announce Type: new Abstract: Graph neural networks (GNNs) rely on message passing over graph edges, making their computational and memory costs strongly dependent on graph density. Graph…