SAE International
PulseAugur coverage of SAE International — every cluster mentioning SAE International across labs, papers, and developer communities, ranked by signal.
6 day(s) with sentiment data
-
Pony.ai aims for 100,000 autonomous trucks by 2030
Pony.ai, a self-driving technology company, has announced ambitious plans to deploy 100,000 light-duty autonomous trucks by 2030, alongside 500 to 1,000 heavy-duty trucks within the next two to three years. This expansi…
-
AI Persona Features Drive Emergent Misalignment, Study Finds
Researchers have identified "persona features" as a key factor in emergent misalignment (EM) in language models, where fine-tuning on a specific task inadvertently leads to harmful behaviors in other areas. Using Sparse…
-
China's new self-driving rules pose major hurdle for Tesla's FSD entry
Tesla is facing significant challenges in gaining approval for its Full Self-Driving (FSD) system in China due to the country's newly established mandatory standards for autonomous driving. These regulations, effective …
-
Self-driving cars use AI and sensors, with 6 automation levels defined
Self-driving cars, also known as autonomous vehicles, use a combination of sensors like lidar, radar, and cameras, along with artificial intelligence and machine learning, to navigate and adjust their speed without huma…
-
New method boosts multilingual LLM performance without retraining
Researchers have developed a novel inference-time method to enhance the performance of multilingual large language models across different languages. This technique, called SAE-Based Steering, utilizes pre-trained spars…
-
Gemma 2-2B research finds active feature planes have less holonomy
A new research paper published on arXiv investigates the concentration of holonomy within specific feature planes of the Gemma 2-2B model. The study preregistered its methodology and analysis rules before inspecting the…
-
AI researchers share experiences publishing on LessWrong and X
Researchers are sharing their experiences publishing AI research on platforms like LessWrong and X (formerly Twitter) to improve communication and foster collaborations. While some have found success in connecting with …
-
LLM agents lose 88% of features via text communication, study finds
A new research paper explores the communication methods of large language model (LLM) agents, specifically investigating whether text-based communication is a bottleneck for complex concept transfer. The study found tha…
-
Global study reveals key drivers for autonomous vehicle acceptance
A new study analyzed data from over 18,000 respondents across 17 countries to understand public acceptance of Level 3 autonomous vehicles. The research, based on the Unified Theory of Acceptance and Use of Technology 2 …
-
AI interpretability training effectiveness debated on LessWrong
A discussion on LessWrong explores the effectiveness of training AI models against interpretability probes. The author argues that such training is only beneficial if the features used by the interpretability methods ar…
-
Model compression minimally impacts Gemma performance, SAEs remain effective
A recent analysis explored the impact of weight compression on Google DeepMind's Gemma 3 4B and Gemma 3 12B models. The study found that performance, measured by cross-entropy and perplexity, remained largely intact eve…
-
LLM refusal research explores distinct harm categories and steering mechanisms
Researchers are investigating the complexities of Large Language Model (LLM) refusal, exploring whether refusal is a distinct concept or intertwined with other training data elements. Experiments with small, open-weight…
-
New AI Research Unveils Methods for Understanding Neural Network Computation
Two new research papers introduce novel methods for understanding the internal workings of complex neural networks. The first, TRACE, proposes a new paradigm for learning to compute on circuit graphs by using a Hierarch…
-
AI model interventions unreliable, new research finds
A new research paper demonstrates that interventions designed to suppress undesirable behaviors in AI models by manipulating Sparse Autoencoder (SAE) features are unreliable. The study shows that even when specific SAE …
-
AI models learn same features but in rotated bases, researchers find
Researchers have discovered that while independently trained transformer models of the same architecture learn similar features, their internal activation representations are rotated by a random amount. This "polymorphi…
-
New protocol detects LLM provider model substitutions
A new research paper proposes a commit-open protocol to detect when hosted large language model providers substitute cheaper models for advertised ones. The protocol uses Merkle trees to commit to sparse autoencoder (SA…
-
WeRide CEO predicts Level 5 driverless cars within 10 years, calling it a 'ChatGPT moment'
WeRide CEO Tony Han predicts that fully autonomous Level 5 (L5) vehicles, capable of driving anywhere under any conditions without human intervention, will become a reality within the next decade. Han likens this potent…