Aégis
PulseAugur coverage of Aégis — every cluster mentioning Aégis across labs, papers, and developer communities, ranked by signal.
5 day(s) with sentiment data
-
New theory explains LLM capability loss during fine-tuning
Researchers have developed a new theoretical framework to explain Pervasive Capability Degradation (PCD) in large language models, a phenomenon where fine-tuning for specific tasks degrades general abilities. The theory…
-
New AEGIS framework protects 3D facial avatar privacy
Researchers have developed AEGIS, a novel framework designed to protect the privacy of 3D facial avatars. This system applies adversarial perturbations to the 3D Gaussian Splatting representation of avatars, effectively…
-
New Lean Prover EasyProver tackles real-world code verification tasks
Researchers have developed EasyProver, a new Lean proving system designed for real-world code verification. This system utilizes recursive agents, isolated contexts, and asynchronous search to tackle complex verificatio…
-
EasyProver system tackles Putnam, IMO benchmarks with recursive agents
Researchers have developed EasyProver, a novel Lean proving system that utilizes recursive agents and asynchronous search to tackle complex mathematical and code verification tasks. This system has demonstrated impressi…
-
New tests proposed to audit LLM failure attribution accuracy
A new paper proposes four tests to distinguish genuine failure attribution in LLMs from spurious correlations with surface cues. These tests, including input restriction and cue following, aim to ensure judges are truly…
-
AEGIS system optimizes deep learning training on shared GPUs
Researchers have developed AEGIS, a runtime scheduling system designed to improve the efficiency of deep learning training on shared multi-GPU servers. AEGIS manages the collocation of multiple deep learning workloads b…
-
New arXiv Paper Questions AI Safety Embedding Methods
A new research paper published on arXiv questions the effectiveness of using a "safe prototype" to determine response safety in AI models. The study found that simply comparing a response's embedding to the average embe…
-
AI agent platform struggles with document utilization, not ingestion
The author details a challenge with their self-hosted AI agent platform, AEGIS, where a vast number of ingested documents, particularly from arXiv, were not being utilized in prompts. Despite storing nearly 90,000 chunk…
-
AEGIS system uses cryptographic attestation for AI decision verification
AEGIS is a system designed to enhance trust in AI by employing cryptographic attestation. It aims to verify the origin, integrity, and authorized status of AI-driven decisions. The system is also being developed with fu…
-
New Aegis technique enhances privacy in medical AI federated learning
Researchers have developed a new privacy-preserving technique called Aegis for federated learning in medical AI. This method aims to protect sensitive patient data by masking gradients during the training process, preve…
-
US explores 'Golden Dome' space missile shield, facing massive cost and strategic risks
The concept of a "Golden Dome" missile defense system, inspired by Ronald Reagan's Strategic Defense Initiative, is being explored by the United States. This ambitious project aims to create a comprehensive space-based …
-
AI infrastructure sees memory innovation, price hikes, and cognitive cost concerns
New developments in AI infrastructure are emerging, with d-Matrix announcing Raptor, a 3D DRAM stack promising high bandwidth at significantly lower power consumption than current High Bandwidth Memory (HBM) standards. …
-
CyberFactory framework trains Aegis model for enhanced cybersecurity LLM capabilities
Researchers have introduced CyberFactory, an open-source framework designed to enhance cybersecurity capabilities in large language models. This framework constructs agentic training data from real-world vulnerabilities…
-
AI community tackles critical security flaws in Model Context Protocol
Security vulnerabilities within the Model Context Protocol (MCP) are being actively discussed and addressed by the AI community. Research indicates a significant percentage of MCP servers lack basic security features li…
-
Digital Foundry Reviews Xenoblade Chronicles 2 on Switch 2
Digital Foundry has released a technical review of Xenoblade Chronicles 2, specifically focusing on its performance on the upcoming Switch 2 console. The review highlights the performance upgrades given to the classic M…
-
AEGIS enhances AI governance with cryptographic assurance
AEGIS has introduced a new system designed to enhance AI governance through cryptographic assurance. This system aims to provide independent verification, ensure long-term integrity of AI systems, and prepare them for p…
-
Security stack with dual LLM approach open-sourced
The developers have open-sourced both halves of their security stack: AEGIS for intrusion detection and ENLIL for LLM-based deliberation. AEGIS is designed for fast, local threat detection, while ENLIL uses a council of…
-
Nobel Laureates Convene at Vatican to Discuss AI and Nuclear War Risks
A recent assembly in Vatican City brought together Nobel laureates, international experts, and religious leaders to discuss the intersection of artificial intelligence and nuclear war. The event, held at the Pope's summ…
-
New AEGIS framework explores span-guided multilingual text detoxification
A new framework called AEGIS has been developed to study the effectiveness of span-level guidance in multilingual text detoxification. This framework combines span detector outputs with frozen generator backbones, allow…
-
New AEGIS system enhances open-source liquid handling robots
Researchers have developed AEGIS, a novel two-layer system designed to enhance the reliability of open-source liquid handling robots like the Opentrons OT-2. The first layer uses a language model to validate protocols a…