AI research

Everything tagged AI research.

Share your story Subscribe

Featured

News

OpenAI puts three AI-generated science claims on separate review tracks

DP News TeamOct 26, 2026

News

Cohere trains reasoning models to reassess confidence after answering

Oct 20, 2026

News

MIT benchmark says GPT-5.5 matched 34.6% of recent theory problems

Oct 18, 2026

Latest

News
Researchers train fixed AI-agent teams to reason together
Researchers developed a framework that learns reusable ways for fixed teams of AI models to divide work and audit answers, outperforming…
DP News Team · 2w ago
News
Meta researchers split control from execution in longer AI agent workflows
Meta researchers proposed an inference-time framework that separates task execution from the control of progress, memory and compute across longer AI…
DP News Team · 3w ago
News
OpenAI releases 722 math manuscripts at varied verification stages
OpenAI has published 722 mathematical manuscripts across 372 result families on GitHub, while warning that the work remains at different stages of…
DP News Team · 3w ago
News
Apple researchers introduce SCLATE for continual-learning agent evaluation
Apple researchers introduced SCLATE, a shared event-scheduling substrate for long-horizon agent benchmarks and post-training.
DP News Team · 4w ago
News
Apple study measures losses when language models serialize structured reasoning
Apple researchers tested how well language models preserve tree-structured expressions when one model turns them into text and another reconstructs…
DP News Team · 4w ago
News
IBM and Marist launch AI incubator with campus IBM z17
IBM and Marist University launched an interdisciplinary AI incubator around an operational campus IBM z17, while its initial projects remain planned…
DP News Team · Sep 29
News
OpenAI forms independent mathematics advisory group with nine researchers
OpenAI announced a nine-member mathematics advisory group. Its unpaid researchers may publish unsolicited advice but cannot make decisions for the…
DP News Team · Sep 21
News
Apple researchers release REVERSAL-BENCH to measure reset-free RL failures
REVERSAL-BENCH varies recoverability across simulated environments to measure when reset-free reinforcement-learning agents become trapped.
DP News Team · Sep 17
News
Google DeepMind launches AlphaGenome Atlas for variant research
Google DeepMind released AlphaGenome Atlas, a queryable one-petabyte catalogue of predicted molecular effects for 9 billion possible single-letter…
DP News Team · Sep 10
News
Google's open Gemma models pass 1 billion downloads and 100,000 variants
Google DeepMind said its open Gemma model family has passed 1 billion cumulative downloads and inspired more than 100,000 community-built variants…
DP News Team · Aug 20
News
Google open-sources HEIR, a compiler that runs AI inference on encrypted data
Google open-sourced HEIR, a compiler that converts pre-trained AI models to run inference directly on encrypted data without decrypting it, the…
DP News Team · Aug 14
News
AI genome model designs 16 viable bacteriophages that kill drug-resistant E. coli
A Stanford and Arc Institute study in Science reports the Evo 2 genome model designed 16 working bacteriophages that kill drug-resistant E. coli.
DP News Team · Aug 6

Featured

OpenAI puts three AI-generated science claims on separate review tracks

Cohere trains reasoning models to reassess confidence after answering

MIT benchmark says GPT-5.5 matched 34.6% of recent theory problems

Latest

Researchers train fixed AI-agent teams to reason together

Meta researchers split control from execution in longer AI agent workflows

OpenAI releases 722 math manuscripts at varied verification stages

Apple researchers introduce SCLATE for continual-learning agent evaluation

Apple study measures losses when language models serialize structured reasoning

IBM and Marist launch AI incubator with campus IBM z17

OpenAI forms independent mathematics advisory group with nine researchers

Apple researchers release REVERSAL-BENCH to measure reset-free RL failures

Google DeepMind launches AlphaGenome Atlas for variant research

Google's open Gemma models pass 1 billion downloads and 100,000 variants

Google open-sources HEIR, a compiler that runs AI inference on encrypted data

AI genome model designs 16 viable bacteriophages that kill drug-resistant E. coli