Research
News in category “Research”
Experiment: Can AI Models Catch All Bugs and Merge External Code into a Flawless Solution?
A practical test demonstrated why developers shouldn't blindly trust a single language model: four LLMs searched for seven bugs in a script, while seven others attempted to assemble a flawless version from external fixes.
Motional and MIT Unveil CW-Net to Make Self-Driving AI Decisions Transparent
Researchers from Motional and MIT have introduced CW-Net, an explainable AI system that translates autonomous vehicle neural network computations into real-time, human-readable concepts.
What role humans will play in the era of advanced AI: takeaways from the ICML conference
At the ICML conference, Professor Arvind Narayanan explained why artificial intelligence will not replace specialists, but will instead shift their core tasks from generating solutions to critically evaluating and overseeing them.
ICML 2026 Highlights: Record Influx of Papers and the Rise of Diffusion Language Models
Seoul hosted one of the premier machine learning conferences, ICML 2026. We break down the event's record statistics, key trends, the triumph of diffusion approaches, and new frontiers in generative AI.
Neural Ranking at VK: How Replacing Gradient Boosting with Neural Networks Boosted Recommendations
The VK AI team shared their experience transitioning from classic decision trees to a multi-task neural network architecture for content ranking, which increased watch time in VK Clips by 5.5%.
A Mathematical Perspective on Gradient Descent: MIT Researchers Introduce Modular Duality Concept
MIT CSAIL researchers have proposed revisiting the classical gradient descent mechanism through the lens of dual vector spaces, bridging theory with Shampoo and µP optimization methods.
How to Make AI Agents Work Reliably: Lessons from Building Maritime Analyst Shippy
Engineers from the Allen Institute for AI shared the architectural challenges encountered while building the Shippy AI agent and explained why language models should not be trusted with executing strict technical algorithms.
MIT Researchers Develop AI to Model Unprecedented Weather Disasters
Engineers at MIT have introduced a machine learning framework capable of generating realistic extreme weather scenarios without relying on past disaster records.
Experiment: Which AI Models Are Named Market Leaders When Asked Themselves
Five popular AI services with web search access answered questions about the best neural network. The test showed that models rarely rank themselves first, but almost always recommend ChatGPT.
Remote Labor Index Benchmark Shows Modern AI Agents Can Handle Only 2.5% of Real Freelance Tasks
Researchers from the Center for AI Safety and Scale AI have developed the RLI benchmark based on hundreds of real Upwork projects. Even leading language models and agents were able to successfully complete no more than 2.5% of the jobs.
Too Good to Be Bad: Study Shows Why AI Struggles to Convincingly Roleplay Villains
Researchers from Tencent and Sun Yat-sen University have found that built-in safety mechanisms prevent large language models from effectively roleplaying negative characters, replacing subtle manipulation with generic aggression.
ClawBench Benchmark Exposes AI Agent Weaknesses: Top Real-Web Success Rate Reaches Just 33%
The new ClawBench benchmark evaluated autonomous web agents on live websites. The best-performing model successfully completed only one-third of the practical tasks.