AI Research

Scientists reveal how language models can manipulate their own alignment process, potentially amplifying harmful biases instead of reducing them.

Researchers introduce a more efficient approach to mapping physical spaces by aligning coordinate frames with the direction of gravity rather than camera position.

New framework treats skills as living assets that evolve across tasks, potentially transforming how language model agents solve problems.

New research reveals how reliance on a single vendor's screening technology produces consistent rejection patterns across demographic groups.

A breakthrough delivery method could accelerate treatments for age-related cognitive decline in humans.

Understand tokens, attention, and next-token prediction without the PhD-level math.

When to customize an LLM: cost, latency, and accuracy tradeoffs explained for engineering leaders.

How RAG fixes LLM hallucinations by letting models fetch real data before generating answers.

A new detection network combining thermal imaging and machine learning aims to prevent collisions as climate-displaced gray whales increasingly stop in busy Bay Area waters.

Most companies want to deploy autonomous AI systems but lack the structural readiness, forcing executives to rethink workflows from the ground up.

A new consolidation technique lets language models compress long conversations into persistent memory, solving a critical scaling bottleneck.

Researchers demonstrate AI can categorize code changes with 84% recall, enabling faster reviews without manual taxonomy engineering.

Researchers bridge vision-language models and segmentation AI to enable instruction-driven instance detection in a single pass.

Researchers combine traditional geometry with neural networks to solve long-standing reconstruction challenges across diverse imaging conditions.

Researchers propose entity-focused approach to keep multimodal models accurate across different video datasets.

Researchers develop a technique to compress powerful video generators into faster versions that work with incomplete information.

Researchers demonstrate that selectively repeating transformer layers in masked diffusion models cuts training costs by 70% while improving reasoning capabilities.

Researchers tackle a fundamental tradeoff in subject-driven synthesis by leveraging multimodal language models alongside specialized identity conditioning.