AI Research

Toward provably private learning from federated data

Mobile Systems

Where the Agent Development Lifecycle Fits

Coordinating agent capability development with the application it powers The post Where the Agent Development Lifecycle Fits appeared first on Towards Data Science.

How to Build a Control Plane for AI Agents

Giving an LLM permission to act, in 9 steps The post How to Build a Control Plane for AI Agents appeared first on Towards Data Science.

😺 48% thought Tavus’s AI was human

Meta’s Muse Hit 5 Million Downloads. The Bigger Story Is How It Got There

Meta’s Muse reached an estimated 5 million downloads in just 22 days, helped by one of the biggest distribution machines in tech. The bigger test is whether Meta can turn that reach into lasting control over how people…

Investigators Found a New Problem With OpenAI’s Rogue Agents: Missing Evidence

OpenAI’s latest rogue-agent story is less about a proven government hack than a growing accountability problem. As agents stitch together outside services, the evidence needed to investigate their behavior can become…

OpenAI Agent Security: The New Rules for Containing AI Agents

An OpenAI Agent Security engineer says frontier AI containment now depends on layered isolation, permissions, monitoring, alignment, and incident-response culture. The Hugging Face incident shows why.

AI Skill of the Day Digest — September 2026 (Part 2)

Explore 13 practical AI skills from September 16–30, with workflows for ChatGPT, Claude, Gemini, Jev, Codex, and AI agents.

Can Light Make Deepfake Detection Cheaper? UCLA Researchers Think So

UCLA researchers are testing a hybrid optical-neural system that screens 15 videos in parallel. The prototype reached 97.79% accuracy on Celeb-DF and could reduce the computing demands of initial deepfake screening.

Characterizing High Bandwidth Flash for LLM Serving

arXiv:2609.39131v1 Announce Type: new Abstract: Large language model (LLM) serving requires substantial memory to store model weights and KV caches. As models grow larger and contexts become longer, memory capacity and…

Simulator-Refined Diffusion for Radio-Frequency Inverse Design

arXiv:2609.38363v1 Announce Type: new Abstract: Diffusion models have shown potential in inverse design of printed circuit boards (PCBs), enabling the generation of layouts conditioned on target S-parameters. Despite…

OpenCollab: A Multi-Agent Coding Framework with Programmable Collaboration and Controllable Runtime

arXiv:2609.38345v1 Announce Type: cross Abstract: Multi-agent coding systems are designed to tackle complex software engineering tasks through collaboration. However, existing evaluations typically assume configured…

From Codebase to Culprit (C2C): Reducing the Search Space for Bugs with Semantic Retrieval and Hierarchical Reinforcement Learning

arXiv:2609.38402v1 Announce Type: cross Abstract: We introduce C2C (From Codebase to Culprit), a framework for precise bug localization that progressively reduces the debugging search space across multiple levels of…

Component-Weighted Centroid Search for Exact Incremental BPE

arXiv:2609.40016v1 Announce Type: cross Abstract: Exact incremental BPE maintains the canonical tokenization state after every appended byte. The recent algorithm of Jiang and Gong (2026) does this in $O(\log^2 t)$…

No Task Vector Is an Island: A Comprehensive Study on the Composability of Task Vectors from On-Policy Distillation

arXiv:2609.39405v1 Announce Type: new Abstract: Task vectors provide a simple mechanism for composing learned capabilities through model merging. However, the composability of task vectors produced by on-policy…

Does a Shared Temperature Imply a Shared Angular Scale in Probabilistic Contrastive Learning?

arXiv:2609.38784v1 Announce Type: new Abstract: In probabilistic contrastive learning, a shared temperature is commonly interpreted as a shared similarity scale, but this interpretation does not hold for…

Self-Evolving Harness on Multiple Tasks with the Agent as Its Own Optimizer

arXiv:2609.38372v1 Announce Type: cross Abstract: A harness is the code around a language-model agent that organizes prompts, calls tools, manages context, and controls execution. As models grow stronger, recent work…

Robust Risk-Sensitive Reinforcement Learning from Corrupted Human Feedback

arXiv:2609.38938v1 Announce Type: new Abstract: Reinforcement learning with human feedback (RLHF) learns from human comparisons, which can be corrupted or deliberately manipulated. This paper studies online…

When Instructions Retrieve Trajectories: Diagnosing and Mitigating Generalization Failures in VLA Models

arXiv:2609.39971v1 Announce Type: cross Abstract: Vision-language-action (VLA) models can exceed 90% success on in-distribution tasks and withstand nuisance changes that preserve the required action, yet fail under…

Explicit Trajectory Diversity for RL-Based Post-Training of LLM Agents

arXiv:2609.38805v1 Announce Type: new Abstract: LLM agents often admit multiple high-quality solutions to the same task, differing in reasoning structure, tool-use pattern, or interaction trajectory. Yet existing…