# May 3, 2025

## Driverless semis have started running regular longhaul routes

- **Driverless trucks** have commenced regular long-haul routes between **Dallas and Houston**, marking a significant milestone in autonomous trucking with Aurora's technology now operating without safety drivers.

## Run LLMs on Apple Neural Engine (ANE)

- **ANEMLL** is an **open-source** project designed to facilitate the **porting of Large Language Models (LLMs)** to the **Apple Neural Engine (ANE)**, enabling on-device inference for enhanced privacy and security in low-power applications.

## Achieving Human Level Competitive Robot Table Tennis

- **Robotic table tennis** has reached **human-level competitiveness**, showcasing advanced AI techniques that enable robots to perform at par with skilled human players, as detailed in the [arXiv paper](https://arxiv.org/abs/2408.03906).

## Open-source AI platform for ear-based sensing applications

- **OpenEarable** is the first fully **open-source AI platform** for ear-based sensing, featuring a modular design with high-precision sensors for diverse applications, including cardiac monitoring and posture tracking.

## Relational Graph Transformers

- **Relational Graph Transformers** revolutionize AI by enabling seamless navigation of relational databases, achieving **20x faster time-to-value** and **30-50% accuracy improvements** through a graph-based approach that eliminates extensive feature engineering.

## [R] Leaderboard Hacking

- **Chatbot Arena rankings are compromised**, as researchers from Stanford and MIT reveal that labs selectively test and release results, leading to a **_biased evaluation_** of LLM performance.

## DuckDB is probably the most important geospatial software of the last decade

- **DuckDB's spatial extension** has revolutionized geospatial data engagement by enabling users to access geo capabilities with just _one or two lines of code_, significantly lowering entry barriers for data generalists.

## [R] Reinforcement Learning for Reasoning in Large Language Models with One Training Example

- **Reinforcement Learning (RL)** enhances **reasoning capabilities** in large language models, enabling effective learning from a single training example, which is a significant advancement in AI efficiency.

## [P] Muyan-TTS: We built an open-source, low-latency, highly customizable TTS model for developers

- **Muyan-TTS** is an **open-source**, low-latency TTS model designed for **easy fine-tuning** and secondary development, primarily supporting English due to limited training data.

## [P] - Deep reinforcement Learning with Unreal Engine

- **UnrealMLAgents** is a new plugin that integrates **Unity ML-Agents** features into **Unreal Engine**, enabling **multi-agent support** and a **Python bridge** for training, enhancing the ease of reinforcement learning setups.