Latest News
Iterative Deployment of LLMs: Rethinking AI Training and Safety
Researchers test iterative deployment as an alternative to reinforcement learning, spotlighting new AI safety risks.
Trellis Cuts Transformer Memory Costs for Long-Context AI Tasks
Trellis uses a fixed-size memory to compress key-value data on the fly, boosting efficiency and accuracy in long-sequence AI challenges.
AI and Legacy Tech Join Forces to Transform Biomedical Innovation
Brain implants, graphene tattoos, and Wi-Fi heart monitoring are reshaping healthcare.
New Learning Framework Boosts Humanoid Robot Coordination
Researchers combine teleoperation and imitation learning to improve humanoid robots’ hand-eye coordination and task performance.
New Two-Stream Deepfake Detector Raises the Bar on Accuracy
Combining semantic encoding with forensic residuals, this method boosts deepfake detection under real-world conditions.
XiaomiMiMo Unveils MiMo-Audio-7B Models, Raising the Bar for Audio AI
MiMo-Audio-7B models push audio processing forward with top-tier performance and few-shot learning capabilities.
Comparing AI Models for Bias Detection: Accuracy and Transparency in Focus
A new study compares transformer-based models for bias detection, revealing how architecture shapes accuracy and reduces false positives.
LLMs Improve Planning by Critiquing Themselves
New research shows large language models boost planning skills by self-review, cutting the need for outside checks.
GASeg Raises the Bar in Self-Supervised Semantic Segmentation
GASeg uses topological methods to improve semantic segmentation, delivering state-of-the-art results on major benchmarks.
LaraCopilot Boosts Laravel Development with Framework-Specific AI Coding
Over 20 teams adopt LaraCopilot, an AI tool that generates production-ready Laravel code, improving efficiency and cutting rework by enforcing architectural standards.
DeepMind and World Labs Want to Prompt Your Next Open World
Google’s AI lab and Fei-Fei Li’s startup are teaming up to automate 3D world-building. It’s a bold bet on 'spatial intelligence' over manual labor.
Stored Household Item Challenge: Raising the Bar for Smarter Service Robots
The Stored Household Item Challenge and NOAM model push service robots closer to human-like reasoning in home environments.
MedKGI Boosts Clinical Diagnosis with AI and Medical Knowledge Graphs
MedKGI combines medical knowledge graphs and structured state tracking to sharpen diagnostic accuracy and speed.
Alibaba's RollArc Boosts AI Training with Disaggregated Systems
RollArc enhances agentic RL efficiency, cutting training times with serverless infrastructure.
Mify-Coder Proves Smaller AI Models Can Outperform Giants in Code Generation
Mify-Coder, a 2.5B-parameter model, beats larger rivals in code generation—while running on everyday hardware.