Aionda

SourceMay 23, 20262026-05-23

Policy Layers for Governing Generalist LLM Agents Safely

How policy-as-code layers can govern generalist LLM agents by controlling tool use, approvals, and data exposure.

RoundupMay 22, 20262026-05-22

AI Resource Roundup (24h) - 2026-05-22

A curated link roundup from recently collected official updates and tech news.

SourceMay 21, 20262026-05-21

Multi-Model LLM Scheduling Under Offloading And Preemption Costs

Examines how offloading and preemption affect multi-model LLM serving under GPU memory limits and model-specific costs.

RoundupMay 20, 20262026-05-20

AI Resource Roundup (24h) - 2026-05-20

A curated link roundup from recently collected official updates and tech news.

COBALT Rethinks Robot Learning Through Smartphone Teleoperation Data

COBALT proposes smartphone and cloud teleoperation to reduce data collection bottlenecks in robot imitation learning.

Limits of Handwritten Math Grading With Vision LLMs

In handwritten math grading, process understanding matters more than OCR, requiring rubric-based review and human checks.

Multi-Image Jailbreaks Expose Multimodal LLM Safety Gaps

Multi-image prompts can bypass single-image filters, exposing structural safety gaps in multimodal LLM defenses.

Neurosymbolic Ternary Claim Verification With Explainable Argumentation Framework

A study on claim verification that proposes ternary decisions and explainable argumentation under incomplete or conflicting evidence.

RoundupApr 4, 20262026-04-04

AI Resource Roundup (24h) - 2026-04-04

A curated link roundup from recently collected official updates and tech news.

RoundupApr 3, 20262026-04-03

AI Resource Roundup (24h) - 2026-04-03

A curated link roundup from recently collected official updates and tech news.

SourceApr 3, 20262026-04-03

Wrapping Florence-2 for ROS 2 Robotic Integration

A case of wrapping Florence-2 with ROS 2 topics, services, and actions for local inference and reproducible integration.

When AI Coding Quality Depends on Task Conditions

AI-generated code quality varies by task and prompt, so security, maintainability, and risk checks matter more than speed alone.

Distributed MADRL Scheduling for Large-Scale Cluster Systems

A look at distributed MADRL for large-scale scheduling, focusing on scalability, adaptability, and design tradeoffs.

Evaluating Harmful Manipulation in Multi-Turn AI Dialogue

A look at research evaluating harmful manipulation through human-AI multi-turn interaction beyond static benchmarks.

How Mathematics Should Govern AI Use Now

Why mathematics must address AI through values, practice, teaching, technology, and ethics to protect autonomy.

Memory and Randomness Bottlenecks in Probabilistic Trustworthy AI

A unified view of probabilistic trustworthy AI: performance bottlenecks may lie in memory and random data movement, not just compute.

NeuroVLM-Bench Tests Clinical Reasoning in Neuroimaging VLMs

A neuroimaging benchmark comparing vision-enabled LLMs on MRI and CT, focusing on clinical reasoning, errors, and safety tradeoffs.

Why Post-Training Collapses Multiple Valid Answers Into One

Examines how LLM post-training collapses multiple valid answers into one and why distributional evaluation matters.

RAG Security Risks From Combined Injection And Poisoning

Examines security risks in RAG when prompt injection and database poisoning combine across retrieval and indexing.

Wireless World Models for AI-Native 6G Networks

How wireless world models combine 3D geometry and wave propagation to improve real-world generalization in AI-native 6G.

RoundupMar 26, 20262026-03-26

AI Resource Roundup (24h) - 2026-03-26

A curated link roundup from recently collected official updates and tech news.

SourceMar 26, 20262026-03-26

Rethinking LLM Agents as Adaptive Computation Graphs

View LLM agents as runtime-adaptive computation graphs to optimize accuracy, cost, latency, debugging, and control.

SourceMar 25, 20262026-03-25

Separating Instructions and Data at the LLM Interface

A look at markup proposals that separate instructions from data in LLM inputs and why structured interfaces matter.